<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Life at Telkomsel - Medium]]></title>
        <description><![CDATA[Our stories to push Indonesia forward through digital technology - Medium]]></description>
        <link>https://medium.com/life-at-telkomsel?source=rss----b17e01250ce9---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Life at Telkomsel - Medium</title>
            <link>https://medium.com/life-at-telkomsel?source=rss----b17e01250ce9---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 08:04:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/life-at-telkomsel" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Building a Secure Enterprise Chatbot Channel with Azure Bot]]></title>
            <link>https://medium.com/life-at-telkomsel/building-a-secure-enterprise-chatbot-channel-with-azure-bot-cbc041b3878f?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/cbc041b3878f</guid>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[microsoft-teams]]></category>
            <category><![CDATA[microsoft-azure]]></category>
            <dc:creator><![CDATA[Putu Adhika]]></dc:creator>
            <pubDate>Fri, 17 Jul 2026 05:45:12 GMT</pubDate>
            <atom:updated>2026-07-17T05:45:11.039Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ab7RiNeJ4rRdYfISCV5qvQ.png" /></figure><p>In a large enterprise such as Telkomsel, automation has become an integral part of daily operations. Various departments have developed their own chatbot-based solutions to streamline processes, improve operational efficiency, and provide quick access to information. These chatbots support a wide range of use cases, from infrastructure monitoring and incident management to application support, reporting, and employee self-service.</p><blockquote>However, as chatbot adoption grows, organizations must also address an important concern: <strong>data security and governance</strong>.</blockquote><p>Historically, many chatbot solutions have relied on third-party messaging platforms such as Telegram because of their ease of use, accessibility, and simple integration capabilities. While these channels are convenient, they may not always align with enterprise security and compliance requirements. In particular, organizations may have limited control over how data is stored, processed, or governed within external platforms that do not have specific agreements, NDAs, or formal partnerships with the company.</p><p>For an enterprise handling critical operational and business information, this introduces challenges around:</p><ul><li>Data security and confidentiality</li><li>Compliance and governance</li><li>Auditability and accountability</li><li>User identity management</li><li>Access control and monitoring</li></ul><p>To address these concerns, Telkomsel has been encouraging the use of communication platforms that are officially sanctioned and supported by the organization.</p><blockquote>The objective is straightforward: ensure that business data remains protected while maintaining clear accountability and governance over enterprise communications.</blockquote><p>One of the most effective approaches is to migrate chatbot interactions from external messaging platforms to Microsoft Teams, a collaboration platform already used internally by employees and governed under enterprise-grade security controls. One of the solution for enabling bot to Microsoft Teams channel is to leverage <strong>Azure Bot Service</strong>.</p><h3>Why Azure Bot Service?</h3><p>Azure Bot Service acts as a bridge between backend chatbot applications and communication channels. Rather than building channel-specific integrations for every messaging platform, developers can build the chatbot logic once and leverage Azure Bot to connect it to supported channels. By positioning Azure Bot between the chatbot engine and Microsoft Teams, organizations can modernize their chatbot ecosystem without significantly modifying the existing business logic.</p><p>The architecture typically consists of the following components:</p><ol><li><strong>Backend Chatbot Application</strong></li></ol><ul><li>Existing bot logic and automation workflows.</li><li>Can be developed using Python, Node.js, .NET, or other supported frameworks.</li></ul><p><strong>2. Azure Bot Service</strong></p><ul><li>Serves as the communication gateway.</li><li>Handles channel connectivity and message routing.</li></ul><p><strong>3. Microsoft Teams Channel</strong></p><ul><li>Provides the user interface where employees interact with the chatbot.</li></ul><p><strong>4. Enterprise Identity &amp; Security Layer</strong></p><ul><li>Microsoft Entra ID authentication.</li><li>Role-based access control.</li><li>Audit and monitoring capabilities.</li></ul><h3>Runbook: Building and Integrating a Chatbot with Teams Using Azure Bot</h3><p>The migration process can be implemented in a series of straightforward steps.</p><p>First, the development team prepares the chatbot application that contains the desired business logic. This chatbot can be a newly developed application or an existing bot previously integrated with Telegram. The application is deployed to Azure App Service, Azure Container Apps, AKS, or another hosting platform with a publicly accessible HTTPS endpoint.</p><p>Once the chatbot service is running, an Azure Bot resource is created within Microsoft Azure. During the provisioning process, the bot is associated with the hosted chatbot endpoint. Azure Bot acts as the middleware responsible for receiving messages from communication channels and forwarding them to the chatbot application.</p><p>After the Azure Bot resource has been configured, developers enable the <strong>Microsoft Teams Channel</strong> from the available Azure Bot channels. This step establishes connectivity between Azure Bot Service and Microsoft Teams without requiring custom Teams-specific development. Microsoft manages the channel integration layer, allowing developers to focus on chatbot functionality rather than communication protocols.</p><p>The next step involves configuring authentication and access controls. Integration with Microsoft Entra ID ensures that only authorized employees can access the chatbot. Organizations may further restrict access based on departments, security groups, or specific Teams environments. This enables the chatbot to leverage enterprise identity management while ensuring compliance with internal security policies.</p><p>Following authentication setup, the chatbot is packaged as a Teams application. Using the Teams Developer Portal or Teams Admin Center, the bot application is registered and published for internal use. Depending on organizational requirements, the bot can be deployed to specific departments, selected users, or the entire enterprise.</p><p>Before production rollout, comprehensive testing should be performed. This includes validating message delivery, user authentication, permission handling, backend connectivity, and monitoring capabilities. Logging and telemetry can be integrated with Azure Application Insights to provide visibility into chatbot performance and usage patterns.</p><p>Once validation is complete, users can interact with the chatbot directly through Microsoft Teams, benefiting from a familiar collaboration environment while the organization maintains control over data security and governance.</p><p>The migration strategy has already demonstrated success within Telkomsel. Several operational chatbot solutions have successfully transitioned from external platforms to Microsoft Teams using Azure Bot Service, including CELYNA, AION, and ARCEUS.</p><h3>Conclusion</h3><p>As enterprises continue to expand automation initiatives, chatbot platforms must evolve beyond convenience and address the growing importance of security, governance, and accountability.</p><p>By migrating chatbot communications from third-party messaging platforms to Microsoft Teams, organizations can ensure that business-critical conversations occur within a controlled and compliant environment. Azure Bot Service provides the missing link by enabling existing chatbot solutions to seamlessly integrate with Teams while maintaining flexibility in backend implementation.</p><p>For Telkomsel, this approach not only strengthens data protection and governance but also provides a scalable foundation for future chatbot adoption across departments.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbc041b3878f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/building-a-secure-enterprise-chatbot-channel-with-azure-bot-cbc041b3878f">Building a Secure Enterprise Chatbot Channel with Azure Bot</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Beyond Cost Savings: Rethinking FinOps (Part 1)]]></title>
            <link>https://medium.com/life-at-telkomsel/beyond-cost-savings-rethinking-finops-part-1-9b7a8a4e97fb?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/9b7a8a4e97fb</guid>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[observability]]></category>
            <category><![CDATA[finops]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[architecture]]></category>
            <dc:creator><![CDATA[Lutfi Ichsan Effendi]]></dc:creator>
            <pubDate>Fri, 10 Jul 2026 04:08:47 GMT</pubDate>
            <atom:updated>2026-07-10T04:30:26.409Z</atom:updated>
            <content:encoded><![CDATA[<h3>We Saved Money. But That is Not the Success Story.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8ShJ-2xOHAo3yw3fKJ9LGg.png" /></figure><p>Cloud cost optimization has become one of the hottest topics in recent years. As cloud adoption continues to grow, organizations are investing in various FinOps practices and tools to gain better control over their cloud spending.</p><p>Most FinOps success stories also sound very similar.</p><p><em>“We reduced our cloud bill by 20%.”</em></p><p><em>“We optimized hundreds of thousands of dollars in cloud spending.”</em></p><p><em>“We eliminated idle resources and improved resource utilization.”</em></p><p>These are impressive achievements.</p><p>But let me ask a simple question.</p><blockquote><strong>Does reducing cloud cost necessarily mean a FinOps initiative is successful?</strong></blockquote><p>Not always.</p><h3>Cost reduction is only the beginning</h3><p>When organizations start implementing FinOps, the first optimizations are usually obvious.</p><p>Unused virtual machines are terminated.</p><p>Idle disks are removed.</p><p>Oversized Kubernetes workloads are right sized.</p><p>Development environments are scheduled to shut down outside working hours.</p><p>Within weeks, the cloud bill starts decreasing.</p><p>Everyone is happy.</p><p>However, several months later, the cloud cost slowly increases again.</p><p>What happened? Did engineers forget everything about optimization?</p><p>Probably not.</p><p>In many cases, the infrastructure was optimized, but the engineering behavior remained exactly the same.</p><p>Teams continued provisioning resources based on assumptions rather than actual utilization.</p><p>New applications/workloads inherited the same resource configurations as previous projects.</p><p>Environments were created without considering long term operational costs.</p><p>Eventually, cloud spending returned to where it was before.</p><p>Because the optimization solved the symptom, not the root cause.</p><h3>Cloud cost is the result of engineering decisions</h3><p>Every cloud bill is simply a collection of engineering decisions.</p><p>Choosing larger instances.</p><p>Allocating more CPU and memory than necessary.</p><p>Keeping environments running 24/7.</p><p>Deploying across multiple Availability Zones.</p><p>Retaining logs longer than required.</p><p>Enable all non required observability metrics.</p><p>None of these decisions are wrong.</p><p>In fact, many of them are absolutely justified depending on the business requirements.</p><p>And the real question is whether the additional cost creates enough value.</p><p>That should be an engineering discussion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LRrSXpLTcAdHarGgGZvkWA.jpeg" /><figcaption>Cloud costs are outputs</figcaption></figure><h3>Visibility creates ownership</h3><p>Imagine receiving the following monthly report.</p><p><strong>Cloud Spend: $1.5 Million</strong></p><p>Useful?</p><p>Yes.</p><p>Actionable?</p><p>Not really.</p><p>Now imagine a different report.</p><ul><li>Service A increased its monthly cost by 40%.</li><li>Kubernetes Cluster B is consistently underutilized.</li><li>Development Environment C runs continuously despite only being used during office hours.</li><li>Network Traffic D has doubled its egress bytes over the past two weeks.</li></ul><p>Suddenly, the conversation changes.</p><p>Instead of asking,</p><p><em>“Why is our cloud bill so high?”</em></p><p>Engineering teams begin asking,</p><p><em>“Why did our service become more expensive?”</em></p><p>That small shift makes a huge difference.</p><p>It becomes another engineering metric that teams actively monitor and improve.</p><h3>FinOps is really about changing engineering behavior</h3><p>One of the biggest values of FinOps is not the dashboard.</p><p>It is not the recommendation engine or the optimization it self.</p><p>Its biggest value is changing how engineers make decisions.</p><p>For example, before deploying a new workload, teams naturally begin asking questions such as:</p><ul><li>Do we really need this much CPU or Storage?</li><li>Is this workload expected to scale?</li><li>Should this environment run all day?</li><li>Are we solving a business problem, or simply over engineering the solution?</li></ul><blockquote>These questions should be appear because engineers have enough visibility to understand the impact of their decisions.</blockquote><h3>A successful FinOps initiative creates a culture</h3><p>Suppose two organizations implement FinOps.</p><p>The first organization reduces cloud spending by 30%.</p><p>Six months later, the cloud bill returns to its previous level because engineering practices never changed.</p><p>The second organization only reduces spending by 10%.</p><p>However, engineering teams now review resource utilization before every production deployment.</p><p>Application owners understand the operational cost of their services.</p><p>Platform teams provide transparent cost visibility across workloads.</p><p>Architecture discussions include cost alongside performance, reliability, and scalability.</p><p>Which organization has the more successful FinOps initiative?</p><p>I would pick the second one.</p><blockquote>Because cloud optimization is an engineering capability that continuously improves as teams make better decisions.</blockquote><h3>Final thoughts</h3><p>FinOps is often measured by a single number:</p><p><em>“How much money did we save?”</em></p><p>While cost reduction is certainly important, I don’t believe it tells the whole story.</p><p>A more meaningful question might be:</p><blockquote><strong>How many better engineering decisions did we make because of FinOps?</strong></blockquote><p>If engineers understand the impact of their infrastructure choices, if application teams take ownership of the services they build and if cost becomes part of everyday engineering discussions, then FinOps has already succeeded, even before the next cloud monthly invoice arrives.</p><p>Because in the end,</p><blockquote><strong>Cloud bills are the consequence of thousands of engineering decisions made every single day.</strong></blockquote><p>In the next article, we will explore why <strong>cloud cost is fundamentally an observability problem </strong>and why <strong>meaningful optimization starts with understanding system behavior rather than simply reducing infrastructure spending.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9b7a8a4e97fb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/beyond-cost-savings-rethinking-finops-part-1-9b7a8a4e97fb">Beyond Cost Savings: Rethinking FinOps (Part 1)</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn Android App Into A Super App Platform In 50 Lines of Code]]></title>
            <link>https://medium.com/life-at-telkomsel/turn-android-app-into-a-super-app-platform-in-50-lines-of-code-8d6370fb7b59?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/8d6370fb7b59</guid>
            <category><![CDATA[telkomsel]]></category>
            <category><![CDATA[mytelkomsel]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[tcmpp]]></category>
            <category><![CDATA[mini-program]]></category>
            <dc:creator><![CDATA[Muhammad Hanif]]></dc:creator>
            <pubDate>Fri, 19 Jun 2026 07:25:14 GMT</pubDate>
            <atom:updated>2026-06-19T07:25:13.059Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kvLMj66ZPou8sUrz479y4Q.jpeg" /></figure><p>Mini programs changed how I think about mobile architecture. No more play store/app store reviews. No more forcing users to install anything. Just open and use.</p><h3><strong>The Day I Realized Our App Was Doing It Wrong</strong></h3><p>Last year, I was sitting in a meeting room when our product owner dropped a bomb: “We need to add 12 new services to the app by next quarter.”<br>Twelve. New. Services.</p><p>I did the math in my head. Each service meant a new module, new dependencies, new tests, new build time, and — most painfully — a new Play Store review cycle for every single bug fix. Our app was already huge. Our users were already complaining about storage. And our build time? Don’t get me started.</p><p>I looked at my colleague and whispered, “<em>There has to be a better way.</em>”</p><p>There was. I just didn’t know it yet.</p><h3>The Rabbit Hole That Changed Everything</h3><p>A few weeks later, I was doom-scrolling tech articles when I stumbled on a <a href="https://skift.com/2025/10/21/grab-expands-beyond-rides-and-food-with-new-partner-apps-for-travel-exclusive/">Skift article about Grab launching “Partner Apps”</a> — third-party services running directly inside the Grab app. HelloRide, redBus, Jolibox — all accessible without additional downloads or logins. Users tapped, used the service, and closed it. The merchant got instant distribution to <a href="https://www.businesswire.com/news/home/20241204105661/en/Superapp-Grab-Selects-AWS-as-its-Preferred-Cloud-Provider-to-Drive-Technology-Innovation-and-Growth">42 million monthly transacting users</a>. Grab got ecosystem stickiness without writing a single line of feature code.</p><p>I kept reading. Then I found WeChat’s numbers.<br><strong>4.5 million mini programs</strong>. Running inside one app. <a href="https://www.yicaiglobal.com/news/wechat-mini-programs-see-13-growth-in-daily-active-users">Serving 450 million daily active users as reported at the WeChat Open Class 2022</a>. By 2025, <a href="https://www.pocketgamer.biz/wechat-mini-games-crosses-500m-monthly-active-users-as-tencent-shifts-focus-to-retention/">WeChat Mini Games alone surpassed 500 million monthly active users</a>. McDonald’s, Nike, government tax filing — all as mini programs. No Play Store. No App Store. Users just tap and use.</p><p>I opened YouTube and watched demos. A user swipes down inside WeChat — a panel of recently used mini programs appears like an app drawer. They tap one. It loads in under a second. Full-featured e-commerce store. Camera access. Payment integration. GPS. The experience felt native, but the user never installed anything.</p><p>Then I looked at Alipay — <strong>4 million more mini programs</strong>. Then LINE in Japan. Zalo in Vietnam. <a href="https://blog.bytebytego.com/p/how-grab-built-an-authentication">Grab across 8 countries and 800+ cities</a>. KakaoTalk in Korea. Even Tencent Cloud had packaged this technology as <a href="https://www.prnewswire.com/apac/news-releases/tencent-cloud-empowers-playsout-unveiling-mini-program-ecosystem-for-global-super-apps-302196646.html">TCMPP (now Tencent Cloud Super App as a Service)</a>, partnering with companies like PlaysOut to bring mini program ecosystems to apps worldwide.</p><p>I sat back and realized this wasn’t some Chinese-only experiment. This was a global architectural shift that had been happening for years — and I’d missed it completely because I was too buried in our own feature backlog.<br>The next morning, our tech lead sent me a message: “<em>Have you looked into mini program SDKs? I think we’ve been solving the wrong problem.</em>” I think this is a coincidence.</p><h3>What Exactly Are Mini Programs?</h3><p>Imagine your phone’s home screen. You have <strong>dozens of apps</strong> — <strong>each one independently installed, independently updated, independently consuming storage</strong>. Now imagine if all those apps could live inside one app. <strong>No installation. Instant loading. Updated silently in the background</strong>.<br>That’s what mini programs are.</p><p>A mini program is a lightweight application that runs inside a host app (the “Super App”) without requiring separate installation.<br>Here’s what makes them different from everything else:</p><ul><li><strong>Unlike native apps </strong>— no Play Store/App Store download required</li><li><strong>Unlike web apps</strong> — they can access native device APIs (camera, GPS, Bluetooth, payments)</li><li><strong>Unlike WebViews</strong> — they have their own rendering engine and lifecycle management</li><li><strong>Unlike PWAs</strong> — they have guaranteed distribution through the host app’s user base</li></ul><p>Think of this way:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yiCSSysCT1EH42EEVe5fUQ.png" /><figcaption>Mini Program Architecture</figcaption></figure><pre>┌─────────────────────────────────────────────────┐<br>│           Your App (The Super App)              │<br>│                                                 │<br>│  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────┐  │<br>│  │ Food    │ │ Finance │ │ Gaming  │ │ ... │  │<br>│  │ Delivery│ │ Service │ │ Mini App│ │     │  │<br>│  └─────────┘ └─────────┘ └─────────┘ └─────┘  │<br>│                                                 │<br>│  ┌─────────────────────────────────────────────┐│<br>│  │     Mini Program Runtime Engine (SDK)       ││<br>│  │     - JavaScript execution                  ││<br>│  │     - Native bridge (device APIs)           ││<br>│  │     - Sandboxed security                    ││<br>│  │     - Lifecycle management                  ││<br>│  └─────────────────────────────────────────────┘│<br>└─────────────────────────────────────────────────┘</pre><p><strong>The super app</strong> provides the <strong>platform</strong>. <strong>Third-party developers </strong>build <strong>mini programs</strong> that run on it. <strong>The host </strong>controls <strong>security, distribution, and user experience</strong>. The mini programs get instant access to millions of users.<br>Everyone wins.</p><h3>The Super App: More Than Just a Buzzword</h3><p>I used to think “Super App” was just marketing speak. Then I looked at the numbers. WeChat in China isn’t just a messaging app. It’s where <a href="https://www.bankmycell.com/blog/number-of-wechat-users/">1.3 billion monthly active users</a> pay bills, order food, book doctors, file taxes, and run businesses — all through mini programs. The average Chinese user opens WeChat 12 times a day and spends 80 minutes inside it.</p><blockquote>But this isn’t just a Chinese phenomenon.</blockquote><p>The super app model is spreading worldwide because it solves a fundamental problem: <strong>users are tired of installing apps</strong>. Research shows that <a href="https://www.mexc.com/news/1107050">77% of users abandon new apps within just three days</a>, and <a href="https://jacksonville.newsnetmedia.com/story/188469/clutch-report-80-of-users-are-forced-to-download-apps/">45% only download an app when they truly need one</a>. Meanwhile, <a href="https://officechai.com/ai/ai-has-led-to-an-explosion-of-new-apps-but-nearly-none-have-managed-to-garner-significant-usage/">AI has led to an explosion of new apps in 2025–2026, but nearly none have managed to garner significant usage</a>. The friction of downloading and installing is a wall that kills conversion. Super apps tear down that wall.</p><h3>The Current State of Mobile Development (And Why It’s Broken)</h3><p>Let me paint you a picture of what “traditional” mobile development looks like in 2024–2025:</p><h4>The Typical Feature Release Cycle</h4><ol><li>Product owner writes requirements (1 week)</li><li>Designer creates mockups (1 week)</li><li>Android and iOS team implements (2–3 weeks)</li><li>QA tests both platforms (1 week)</li><li>Submit to Play Store and App Store (1–3 days review)</li><li>Users update their app (30% never do)</li></ol><p>Total time from idea to user’s hands: <strong>4–6 weeks minimum</strong>.<br>Now imagine a bug is found. Go back to step 3. <strong>Another 1–2 weeks</strong>.</p><h4>The Numbers That Keep CTOs Awake</h4><pre><br>+-----------------------------------+---------------------------------------+<br>|            Pain Point             |                Reality                |<br>+-----------------------------------+---------------------------------------+<br>| Average app size                  | 20–50MB (growing every year)          |<br>| User retention after 30 days      | 5.7% (95% of users abandon your app)  |<br>| App store review time             | 1–7 days (unpredictable)              |<br>| Cost of maintaining 2 codebases   | 2x developers, 2x bugs, 2x headaches  |<br>| Monthly app downloads per user    | 0 (most users don&#39;t install new apps) |<br>| Build time for large Android apps | 15–45 minutes                         |<br>+-----------------------------------+---------------------------------------+</pre><h4><strong>The Mini Program Alternative</strong></h4><ol><li>Developer builds feature using web tech (1–2 weeks)</li><li>Deploy via OTA — no store review (minutes)</li><li>Available to all users instantly (no update needed)</li><li>Runs on both Android and iOS from one codebase</li><li>App size stays the same (mini programs load on-demand)</li></ol><p>Time from idea to user’s hands: <strong>2–3 weeks</strong>. Bug fix: <strong>hours, not weeks</strong>.<br>This is why companies are shifting. It’s not ideology — it’s economics.</p><h3>The Honest Truth: Pros and Cons</h3><p>I’m not here to tell you snake oil. Mini programs aren’t perfect. Here’s my opinions after working with them:</p><h4>Why Mini Programs Win</h4><ol><li><strong>Zero Installation Friction</strong><br>This is the killer feature. I’ve seen conversion rates jump 3–5x when services moved from “download our app” to “just open this mini program.” In emerging markets like Indonesia where storage is precious, this is a game-changer.</li><li><strong>Instant Updates<br></strong>Remember when Log4j happened and everyone panicked? With mini programs, you push a fix in minutes. No store review. No waiting for users to update. The next time they open the mini program, they get the latest version automatically.</li><li><strong>One Codebase, All Platforms<br></strong>Mini programs are built with JavaScript/Typescript/HTML/CSS-like frameworks. Write once, run on Android and iOS. Your team just got twice as productive.</li><li><strong>Leverage Existing Users<br></strong>If your app has 10 million users, every mini program you host has instant access to those 10 million users. No marketing budget needed for distribution.</li><li><strong>Ecosystem Growth Without Engineering Debt<br></strong>Want to add 50 new services? You don’t need 50 new feature teams. Third-party partners build their own mini programs. You just provide the platform.</li><li><strong>Smaller App Size<br></strong>Your base app stays lean. Services load on-demand as lightweight packages. Users in bandwidth-constrained areas will thank you.</li></ol><h4>Where Mini Programs Fall Short</h4><ol><li><strong>Performance Ceiling</strong><br>For computationally <strong>heavy tasks (3D games, video editing, real-time AR), native still wins</strong>. Mini programs add a layer of abstraction that costs ~<strong>10–15% performance overhead</strong>.</li><li><strong>Platform Dependency</strong><br>You’re building on someone else’s runtime. If the SDK provider changes their API, deprecates a feature, or goes out of business — you’re affected.</li><li><strong>Limited Deep Hardware Access</strong><br>Need Bluetooth Low Energy with custom GATT profiles? Need to process on-device ML models? Need background services running for hours? Native is still your answer.</li><li><strong>Fragmentation</strong><br>WeChat mini programs, Alipay mini programs, LINE mini apps, TCMPP — they’re all slightly different. There’s no universal standard (yet). Code portability between platforms is limited.</li><li><strong>Revenue Sharing</strong><br>Some platforms take a cut of transactions happening inside mini programs. You need to factor this into your business model.</li><li><strong>Debugging Complexity</strong><br>When something goes wrong, the bug could be in your mini program code, the SDK runtime, the bridge layer, or the host app’s configuration. The debugging surface area is larger.</li></ol><h4>My Verdict</h4><pre><br>+------------------------------------+-----------------------------+<br>|              Usecase               |       Recommendation        |<br>+------------------------------------+-----------------------------+<br>| E-commerce, marketplaces           | ✅ Mini program (no-brainer) |<br>| Utilities &amp; tools                  | ✅ Mini program              |<br>| Content &amp; media                    | ✅ Mini program              |<br>| Casual games                       | ✅ Mini program              |<br>| Heavy computation (ML, video)      | ❌ Keep native               |<br>| IoT / hardware integration         | ❌ Keep native               |<br>| Mission-critical real-time systems | ❌ Keep native               |<br>+------------------------------------+-----------------------------+</pre><blockquote>The sweet spot: use mini programs for your ecosystem features, keep native for your core platform.</blockquote><h3>Let’s Build It!</h3><p>Let me show you how I integrated mini program capabilities into an Android app. I’m using Tencent Cloud Super App as a Service (formerly known as TCMPP — Tencent Cloud Mini Program Platform).<br>Why TCMPP? Because it gives you the same mini program runtime that powers WeChat’s ecosystem, but as a standalone SDK you can embed in any app. It’s battle-tested at WeChat scale (1.3 billion users) and <a href="https://intl.cloud.tencent.com/document/product/1219/68180">now available as a cloud service (renamed to Tencent Cloud Super App as a Service in January 2025)</a>.</p><h4>What We’re Building</h4><p>A simple Android app that can:</p><ul><li>Initialize the mini program runtime</li><li>Open any mini program by its ID</li><li>Provide native capabilities (image loading, permissions) to mini programs</li></ul><p>Prerequisites:</p><ul><li>Android Studio (latest)</li><li>An Android project with minSdkVersion 21+</li><li>A Tencent Cloud account (for the configuration file)</li><li>~15 minutes of your time</li></ul><p><strong>Step 1: Add the Maven Repository</strong></p><p>Open settings.gradle and add the TCMPP repository:</p><pre>// settings.gradle<br>pluginManagement {<br>    repositories {<br>        google()<br>        mavenCentral()<br>        maven {<br>            url &#39;https://maven-dev.tcmppcloud.com/fHKFBbEjd/repository/maven-public/&#39;<br>        }<br>    }<br>}<br><br>dependencyResolutionManagement {<br>    repositoriesMode.set(RepositoriesMode.FAIL_ON_PROJECT_REPOS)<br>    repositories {<br>        google()<br>        mavenCentral()<br>        maven {<br>            url &#39;https://maven-dev.tcmppcloud.com/fHKFBbEjd/repository/maven-public/&#39;<br>        }<br>    }<br>}</pre><p><strong>Step 2: Add Dependencies</strong></p><p>In your app module’s build.gradle:</p><pre>// app/build.gradle<br>plugins {<br>    id &#39;com.android.application&#39;<br>    id &#39;org.jetbrains.kotlin.android&#39;<br>    id &#39;org.jetbrains.kotlin.kapt&#39;<br>}<br><br>android {<br>    defaultConfig {<br>        minSdk 21<br>        // ... your existing config<br><br>        packagingOptions {<br>            pickFirst &#39;lib/arm64-v8a/libc++_shared.so&#39;<br>            pickFirst &#39;lib/armeabi/libc++_shared.so&#39;<br>            pickFirst &#39;lib/armeabi-v7a/libc++_shared.so&#39;<br>            pickFirst &#39;lib/arm64-v8a/libwechatxlog.so&#39;<br>            pickFirst &#39;lib/armeabi/libwechatxlog.so&#39;<br>            pickFirst &#39;lib/armeabi-v7a/libwechatxlog.so&#39;<br>        }<br>    }<br>}<br><br>dependencies {<br>    // Required support libraries<br>    implementation &#39;com.google.android.material:material:1.3.0-alpha03&#39;<br>    implementation &#39;androidx.core:core-ktx:1.6.0&#39;<br>    implementation &#39;com.google.code.gson:gson:2.8.6&#39;<br>    implementation &#39;com.squareup.okhttp3:okhttp:3.12.13&#39;<br><br>    // 🔑 TCMPP Mini Program SDK — the magic<br>    kapt &#39;com.tencent.tcmpp.android:mini_annotation_processor:2.0.0&#39;<br>    implementation &#39;com.tencent.tcmpp.android:mini_core:2.0.0&#39;<br>    implementation &#39;com.tencent.tcmpp.android:mini_baselib:2.0.0&#39;<br>}</pre><p><em>💡 Check the </em><a href="https://www.tencentcloud.com/document/product/1219/61823"><em>latest SDK version</em></a><em> and replace 2.0.0 accordingly.</em></p><p><strong>Step 3: Get Your Configuration File</strong></p><p>This is the one step that requires the Tencent Cloud console:</p><ol><li>Go to <a href="https://www.tencentcloud.com/account/login?s_url=https%3A%2F%2Fconsole.tencentcloud.com%2F">Tencent Cloud Super App as a Service Console</a></li><li>Click “Create Superapp”</li><li>Enter your app name and applicationId (must match your build.gradle)</li><li>Download tcsas-android-configurations.json</li><li>Drop it into app/src/main/assets/</li></ol><pre>app/src/main/assets/<br>└── tcsas-android-configurations.json   ← Your SDK config</pre><p>This file contains your app’s identity, API keys, and domain configurations. Without it, the SDK won’t initialize.</p><p><strong>Step 4: Cache Your Application Instance</strong></p><p>The TCMPP SDK uses multi-process architecture (mini programs run in a separate process for isolation). This means your Application instance must be accessible across processes:</p><pre>// MyApp.kt<br>class MyApp : Application() {<br><br>    companion object {<br>        lateinit var instance: MyApp<br>            private set<br>    }<br><br>    override fun attachBaseContext(base: Context) {<br>        super.attachBaseContext(base)<br>        instance = this  // Cache early — before any ContentProvider runs<br>    }<br>}</pre><p>Don’t forget to register it in AndroidManifest.xml:</p><pre>&lt;application<br>    android:name=&quot;.MyApp&quot;<br>    ... &gt;</pre><p><strong>Step 5: Implement MiniConfigProxy (The Core Integration)</strong></p><p>This is where the SDK knows how to initialize. It’s the bridge between your app and the mini program runtime:</p><pre>// MiniConfigProxyImpl.kt<br>import android.app.Application<br>import com.tencent.tcmpp.android.api.MiniConfigProxy<br>import com.tencent.tcmpp.android.api.MiniInitConfig<br>import com.tencent.tcmpp.android.api.annotations.ProxyService<br><br>@ProxyService(proxy = MiniConfigProxy::class)<br>class MiniConfigProxyImpl : MiniConfigProxy() {<br><br>    override fun getApp(): Application = MyApp.instance<br><br>    override fun buildConfig(): MiniInitConfig {<br>        return MiniInitConfig.Builder()<br>            .configAssetName(&quot;tcsas-android-configurations.json&quot;)<br>            .autoRequestPermission(true)<br>            .debug(BuildConfig.DEBUG)<br>            .build()<br>    }<br>}</pre><p>That’s it. The <a href="http://twitter.com/ProxyService">@ProxyService</a> annotation tells the SDK to use this class via compile-time code generation (that’s what the kapt processor does).</p><p><strong>Step 6: Launch a Mini Program</strong></p><pre>// MainActivity.kt<br>import com.tencent.tcmpp.android.api.TmfMiniSDK<br>import com.tencent.tcmpp.android.api.model.MiniStartOptions<br><br>class MainActivity : AppCompatActivity() {<br><br>    override fun onCreate(savedInstanceState: Bundle?) {<br>        super.onCreate(savedInstanceState)<br>        setContentView(R.layout.activity_main)<br><br>        findViewById&lt;Button&gt;(R.id.btnOpenMiniApp).setOnClickListener {<br>            // One line. That&#39;s all it takes.<br>            TmfMiniSDK.startMiniApp(this, &quot;your-mini-app-id&quot;, MiniStartOptions())<br>        }<br>    }<br>}</pre><p>And that’s it. Your app is now a super app platform. 🎉</p><p><strong>Step 7 (Bonus): Implement Image Loading</strong></p><p>The SDK intentionally doesn’t bundle an image loader (to avoid conflicts with whatever you’re already using). Add this so mini programs can display images:</p><pre>// MiniImageProxyImpl.kt<br>@ProxyService(proxy = MiniImageProxy::class)<br>class MiniImageProxyImpl : MiniImageProxy() {<br><br>    override fun loadImage(imageView: ImageView, url: String, options: MiniImageOptions?) {<br>        Glide.with(imageView.context)<br>            .load(url)<br>            .placeholder(R.drawable.ic_placeholder)<br>            .into(imageView)<br>    }<br>}</pre><p><strong>The Final File Structure</strong></p><pre>app/<br>├── build.gradle<br>├── src/main/<br>│   ├── AndroidManifest.xml<br>│   ├── assets/<br>│   │   └── tcsas-android-configurations.json<br>│   └── kotlin/com/yourapp/<br>│       ├── MyApp.kt                   (5 lines of real code)<br>│       ├── MiniConfigProxyImpl.kt     (12 lines of real code)<br>│       ├── MiniImageProxyImpl.kt      (8 lines of real code)<br>│       └── MainActivity.kt           (1 line to open mini program)<br>settings.gradle                        (add maven URL)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*6U7WxP1Y8aI-n6rNjCnrTw.gif" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oJqwjIy2srBcYwAB0twH3A.png" /><figcaption>left: mini program sample app, right: world cup mini program developed by Telkomsel</figcaption></figure><p>Total integration code: <strong>~50 lines</strong>. As promised.</p><h3>What Happens Next</h3><p>Once your app is a mini program host, the possibilities explode:</p><ul><li>Partners build services for YOUR platform — they bring their own engineering resources.</li><li>You control the ecosystem — approve/reject mini programs, set guidelines, manage quality.</li><li>Users never leave your app — every service they need is inside your app.</li><li>You ship features without app updates — deploy new mini programs via the console, not the Play Store.</li><li>You measure everything — the SDK gives you analytics on mini program usage, retention, crashes</li></ul><p>This is exactly what MyTelkomsel did in 2024. It went from a telecom self-service app to a digital lifestyle platform. Gaming? Mini program. Entertainment? Mini program. Financial services? Mini program. All without bloating its APK or slowing core experience.</p><h3>Lessons Learned (The Hard Way)</h3><p>After months of working with mini programs, here are things I wish someone told me upfront:</p><ol><li><strong>Package name mismatch will haunt you</strong>. If your applicationId in build.gradle doesn’t exactly match what’s configured in the console, the SDK will silently fail. No crash. No error toast. Just… nothing happens. I lost much hours to this.</li><li><strong>Multi-process means multi-headache (initially)</strong>. Mini programs run in a separate process. This means singletons, shared preferences, and in-memory caches in your main process are NOT accessible from mini programs. Plan your data sharing strategy upfront.</li><li><strong>Start with the demo app</strong>. Before integrating into your production app, clone the <a href="https://github.com/TencentCloud/tcmpp-demo-android">TCMPP demo on GitHub</a> and get it running. Understand the lifecycle. Then integrate.</li><li><strong>Mini programs are NOT WebViews</strong>. I made this mistake in my mental model. They have their own rendering engine, their own JS runtime, and their own component system. Treat them as first-class citizens, not embedded web pages.</li><li><strong>Think platform-first</strong>. The real value isn’t in building mini programs yourself — it’s in enabling others to build them for your platform. Create developer documentation, review processes, and quality guidelines early.</li></ol><h3>The Bigger Picture</h3><p>We’re witnessing a fundamental shift in mobile architecture. <strong>The monolithic app is dying — not because it’s bad, but because it doesn’t scale to meet user expectations</strong>.</p><p>Users want one app that does everything. Developers want to ship fast without store reviews. Product owners want to onboard partners without engineering bottlenecks. Businesses want to grow their ecosystem without proportional cost growth.</p><p>Mini programs solve all of this.</p><p>In 5 years, I believe every major consumer app will be a platform. The question for you is: <strong>will you be the platform, or will you be the mini program running on someone else’s?</strong></p><ul><li><a href="https://medium.com/life-at-telkomsel/code-smell-dispensable-3e2ab499d0f1">Code Smell — Dispensable</a></li><li><a href="https://medium.com/life-at-telkomsel/aws-cloud-quest-gamification-for-training-and-certification-bc9d7ec858e8">AWS Cloud Quest: Gamification for Training and Certification</a></li><li><a href="https://hanifmhd.medium.com/how-secure-is-your-mobile-app-part-1-deb067c0fa4b">How Secure is Your Mobile App? — Part 1</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8d6370fb7b59" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/turn-android-app-into-a-super-app-platform-in-50-lines-of-code-8d6370fb7b59">Turn Android App Into A Super App Platform In 50 Lines of Code</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quantum Threat Wake-Up Call for Indonesian Digital Organizations: Start the PQC Roadmap Now]]></title>
            <link>https://medium.com/life-at-telkomsel/quantum-threat-wake-up-call-for-indonesian-digital-organizations-start-the-pqc-roadmap-now-7af251de06a8?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/7af251de06a8</guid>
            <category><![CDATA[quantum-computing]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Febriy]]></dc:creator>
            <pubDate>Tue, 10 Feb 2026 02:39:19 GMT</pubDate>
            <atom:updated>2026-02-10T02:39:18.024Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zfu3DWZcBZcPRcjzkY2rsw.jpeg" /><figcaption>Cryptography Illustration</figcaption></figure><p>History of cryptography dates back thousands of years. One of the earliest known examples comes from a clay tablet in Mesopotamian era, which used enciphered characters to conceal ingredients of ceramic glazes — an early attempt to protect intellectual property. Later, in Ancient Rome around 100 B.C., Julius Caesar employed a simple substitution technique known as Caesar cipher, which replaced each letter with another positioned a few places away in alphabet. This method represents one of the first documented uses of cryptography for securing personal and sensitive communications.</p><p>In 1917, Edward Hebern introduced a groundbreaking concept by combining electrical circuits with typewriter components to create first rotor-based encryption machine while he was in prison. This device allowed messages to be encrypted automatically as they were typed, and decryption required by retyping the ciphertext and reversing the rotor. Building on Hebern’s idea after World War I, German cryptologist Arthur Scherbius developed Enigma machine, which became a critical tool for encrypting top-secret communications during World War II. Enigma’s security relied on precise calibration and unique keys, making it extremely challenging to break. However, a major turning point came when Alan Turing and his team, working alongside Polish cryptanalysts, engineered a machine capable of decrypting Enigma messages — a milestone that significantly contributed to the Allied to support war strategy.</p><p>The next big step was <strong>public-key cryptography</strong>, which solved a big problem: how to share information securely without first agreeing on a secret key. In 1977, <strong>RSA algorithm</strong> was introduced. RSA uses two keys:</p><ul><li>A <strong>public key</strong> to encrypt data.</li><li>A <strong>private key</strong> to decrypt it.</li></ul><p>This made secure communication possible over internet and enabled things like digital signatures and identity verification. Later, in 1990s, <strong>Elliptic Curve Cryptography (ECC)</strong> was developed. ECC did same job as RSA but with much smaller keys, making it faster and more efficient — perfect for mobile devices and IoT.</p><p>Today, RSA and ECC are everywhere: securing websites (TLS), emails, VPNs, software updates, and more. But there’s a new challenge: <strong>quantum computers</strong>. They could break math behind RSA and ECC, which is why cryptography is now moving toward new, Post-Quantum Cryptography.</p><h3><strong>Quantum Era</strong></h3><p>Cryptographic algorithms are the most important safeguard for confidentiality of electronic information from unauthorized access. For decades, these algorithms have proved strong enough to defend against attack using conventional computers that attempt to defeat cryptography. However, future quantum computing may be able to break these algorithms, rendering data and information vulnerable. To address the potential issue requires <strong>new cryptographic methods</strong> that can <strong>protect data from </strong>both<strong> current conventional computers </strong>and <strong>quantum computers</strong>. These methods are named <strong>post-quantum cryptography (PQC)</strong>.</p><p>Modern encryption stays safe mainly because it is built on math problems that are extremely slow to solve on ordinary (classical) computers. Quantum computers change game for some of these problems because they can use quantum effects (superposition and interference) to extract hidden structure from math, rather than checking possibilities one by one.</p><p>For RSA, the core “lock” is that it’s easy to multiply two huge prime numbers to get a <em>safe-number (N = p </em>x <em>q</em>), but (for classical computers) very hard to reverse that and recover p and q. Shor’s algorithm shows that a sufficiently capable quantum computer could factor N efficiently by turning factoring into a period-finding problem and using a quantum routine (often described with Quantum Fourier Transform as a key ingredient) to learn that period; once the period is known, prime numbers can be computed, and RSA private key can be reconstructed.</p><p>For Diffie–Hellman and elliptic-curve cryptography (ECC), the “lock” is discrete logarithm problem: you can easily compute a public value from a secret number, but it’s infeasible for classical computers to recover the secret from public value at standard key sizes. Shor’s approach also applies here (it covers discrete logs as well as factoring), which is why RSA, DH, and ECC are commonly categorized as quantum-vulnerable public-key cryptography.</p><p>For symmetric encryption like AES, quantum computers do not have a known method that “breaks” AES like the way Shor breaks RSA/ECC. The main impact is Grover’s algorithm, which gives a quadratic speedup for brute-force key search. In practical terms, this is why increasing symmetric key sizes (e.g., preferring 256-bit keys when appropriate) is considered a straightforward mitigation.</p><p>At the system level, a widely discussed risk is “harvest now, decrypt later”: an attacker can record encrypted traffic today and hope to decrypt it in the future once large fault-tolerant quantum computers exist. NIST explicitly highlights this as a driver for starting migration to post-quantum cryptography (PQC), and NSA’s CNSA 2.0 guidance signals that RSA/DH/ECC will be deprecated under mandates and points toward quantum-resistant replacements on a transition timeline.</p><p>Currently, PQC does not rely on just one standard, there are several standards and guidance created to support digital security ecosystem after enablement of quantum computing. PQC standards fulfill several essential roles: they identify and approve quantum-resistant algorithms, outline best practices for integrating these algorithms into real systems, and establish testing procedures, compliance framework, and policy directives to ensure national security environments.</p><p>Historically, journey from algorithm standardization to full integration into information systems can take 10 to 20 years. This timeline reflects the complexity of organizations building algorithms into products and services, procuring those products and services, and integrating those products and services into technology infrastructures.</p><p>According to Consumer Electronics Show 2026 discussion (reported by Xinhuanet), Experts mentioned that the gap between current quantum capabilities and business needs has narrowed significantly. They also warned that adoption may accelerate faster than expected like the Artificial intelligence moment when the ChatGPT released.</p><p><strong>So, what should we do now?</strong></p><p>We cannot simply wait for quantum computing to launch widely; proactive preparation is required. Organizations must begin migrating to Post-Quantum Cryptography (PQC) now to prevent future breaches, because attackers are already employing “harvest now, decrypt later” strategy — collecting encrypted data today with intent to decrypt it once quantum capabilities available. Early adoption of PQC ensures that sensitive information remains secure even against tomorrow’s quantum threats.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uPqjIsKn0a5YfTud9T2Ilw.png" /><figcaption>NCSC Timeline for PQC Migration</figcaption></figure><p>According to guidance from United Kingdom’s National Cyber Security Centre (NCSC), migration to Post-Quantum Cryptography (PQC) is planned to begin in early 2026 and conclude by 2035. The transition is organized into three key milestones:</p><ul><li>By 2028: Organizations should have established clear migration objectives, completed a thorough cryptographic assessment, and developed an initial migration plan.</li><li>By 2031: Migration of the highest-priority and most critical systems should be completed.</li><li>By 2035: All systems, services, and products across the organization should be fully migrated to PQC.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/757/1*OBCb4V8V8Heot8ANzGkMtA.png" /><figcaption>PQC Migration Proposal Timeline for Indonesian Digital Organization</figcaption></figure><p>For Indonesian organizations, awareness of Post-Quantum Cryptography (PQC) remains relatively low, making early education and preparation is critical. To address this, we propose dividing migration timeline into five milestones while remaining aligned with global guidelines from NIST and UK’s NCSC. First milestone, beginning in 2026, focuses on building organizational awareness of PQC. This includes executive briefings to present quantum threat landscape and NIST-approved PQC standards to leadership, conducting workshops for IT, security, and development teams, and engaging vendors to understand their PQC readiness and roadmap. In parallel, organizations should perform an inventory assessment of cryptographic assets (including certificates, keys, algorithms, and protocols) and ensure alignment with government regulations, particularly for sensitive data that requires priority protection.</p><p>Second milestone, started in 2027, to initiate migration plan and define a strategy to ensure business continuity and minimize disruption to critical services. This phase should include risk assessments, prioritization of systems based on sensitivity and exposure, and evaluation of hybrid cryptographic approaches to maintain interoperability during the transition. By 2028, organizations should complete initial planning process and begin migrating high-priority and critical services, especially those that store or transmit sensitive data. This phase marks the start of practical implementation and requires close coordination between internal teams and technology vendors to ensure smooth integration.</p><p>Third milestone, targeted for 2031, focuses on completing migration for all high-priority systems, ensuring that most critical components of the organization infrastructure are quantum-safe. Finally, by 2035, the organization should achieve full adoption of PQC across all systems, services, and products. This comprehensive migration will safeguard long-term data confidentiality and integrity against quantum threats, positioning organizations to remain compliant with quantum standards and resilient in the face of emerging cybersecurity challenges.</p><p><strong>How to measure success on PQC adoption?</strong></p><p>To evaluate success adoption of Post-Quantum Cryptography (PQC), organizations should focus on three critical aspects: <strong>System Implementation</strong>, <strong>Security Effectiveness, and Operational Impact</strong>. <strong>System Implementation</strong> measures extent of cryptographic algorithm implementation, system updates, and ensuring that PQC solutions are properly embedded into existing infrastructure. <strong>Security Effectiveness</strong> assesses the strength of cryptographic protections by monitoring vulnerability exposure and compliance with recognized standards including standards from international organizations and governments. <strong>Operational Impact</strong> examines how migration affects business operations, including user experience and any modifications to core system processes. These 3 aspects provide a comprehensive view of progress and readiness for Post-Quantum Cryptography adoption.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IWa-6XL8eajHxO8VI428xA.png" /><figcaption>Critical Aspects to Measure PQC Success</figcaption></figure><p>While development of quantum computing is still evolving, our organization should proactively prepare for security risks that may emerge as quantum capabilities become more widely available. In particular, we must take action now to reduce exposure to “harvest now, decrypt later” threats, where adversaries collect encrypted data today with the intent to decrypt it in the future. By strengthening our cryptographic posture, prioritizing quantum-resistant approaches, and establishing a clear transition roadmap, we can mitigate long-term risk and protect sensitive information throughout and beyond the quantum era.</p><h4><strong>References:</strong></h4><ol><li><a href="https://www.entrust.com/resources/learn/history-of-cryptography">The History of Cryptography: Timeline &amp; Overview | Entrust</a></li><li><a href="https://www.ibm.com/think/topics/cryptography-history">The History of Cryptography | IBM</a></li><li><a href="https://nvlpubs.nist.gov/nistpubs/ir/2024/NIST.IR.8547.ipd.pdf">NIST IR 8547 initial public draft, Transition to Post-Quantum Cryptography Standards</a></li><li><a href="https://arxiv.org/abs/quant-ph/9508027">https://arxiv.org/abs/quant-ph/9508027</a></li><li><a href="https://arxiv.org/abs/quant-ph/9605043">https://arxiv.org/abs/quant-ph/9605043</a></li><li><a href="https://www.ncsc.gov.uk/guidance/pqc-migration-timelines#section_4">Timelines for migration to post-quantum cryptography — NCSC.GOV.UK</a></li><li><a href="https://wqs.events/metrics-dashboard-tracking-your-pqc-migration-kpis-for-quantum-readiness/#essential-kpis-for-pqc-migration">wqs.events/metrics-dashboard-tracking-your-pqc-migration-kpis-for-quantum-readiness/#essential-kpis-for-pqc-migration</a></li><li><a href="http://www.china.org.cn/world/Off_the_Wire/2026-01/10/content_118273024.shtml">Feature: Quantum computing nears mainstream as experts predict “ChatGPT moment”</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7af251de06a8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/quantum-threat-wake-up-call-for-indonesian-digital-organizations-start-the-pqc-roadmap-now-7af251de06a8">Quantum Threat Wake-Up Call for Indonesian Digital Organizations: Start the PQC Roadmap Now</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Our Organization Secures Production Environment with ELEVATE Temporary Elevated Access Management]]></title>
            <link>https://medium.com/life-at-telkomsel/our-organization-secures-production-environment-with-elevate-temporary-elevated-access-management-7cb7b68cac55?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/7cb7b68cac55</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[design-systems]]></category>
            <category><![CDATA[design-thinking]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <dc:creator><![CDATA[Dwiki Kurnia]]></dc:creator>
            <pubDate>Mon, 22 Dec 2025 08:32:51 GMT</pubDate>
            <atom:updated>2025-12-22T08:32:39.440Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*soh-YTVP21w3vbuRgkrAyg.png" /></figure><p>Keeping our production environment safe is a top priority. In the past, giving team members permanent high-level access made things easier, but it also opened the door to mistakes and security risks. We needed a better way, one that’s both secure and simple. That’s why we introduced <strong>ELEVATE</strong>, a system that gives temporary, just-in-time access only when it’s really needed.</p><h3>What is ELEVATE?</h3><p>ELEVATE is our answer to the challenge of balancing security and productivity. Instead of handing out permanent admin rights, ELEVATE lets team members request higher access for a short time. Once their task is done, the extra access is automatically taken away. This keeps our environment safe and our team agile.</p><p>Before ELEVATE, we faced a few big problems:</p><ul><li><strong>Permanent admin access</strong> meant more chances for mistakes or security issues.</li><li><strong>Higher security risk w</strong>ith more people holding powerful permissions all the time, our production environment was more exposed to potential threats.</li><li><strong>Complex user management</strong> made it hard to keep track of who had access to what.</li></ul><p>We wanted a solution that was fast, safe, and easy to use. ELEVATE was designed to solve these problems by making access temporary and approvals quick.</p><h3>How ELEVATE Works (Step by Step)</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rILK2UziXc1Lhx23UkqsfQ.png" /><figcaption>ELEVATE Architecture and End-to-end Process</figcaption></figure><p>Here’s how ELEVATE keeps things simple and secure:</p><ol><li><strong>Request Access</strong><br>Team members go to our AWS access portal, select ELEVATE, and fill out a quick form to request higher access.</li><li><strong>Approval</strong><br>The request goes to a manager or approver, who checks and approves it with just a click.</li><li><strong>Temporary Access</strong><br>Once approved, the team member gets elevated access for a limited time, just enough to finish the job.</li><li><strong>Automatic Removal</strong><br>When the time is up, ELEVATE automatically removes the extra access. No need to remember to turn it off.</li><li><strong>Full Tracking</strong><br>Every request and action is logged, so we always know who did what and when.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*asGbZHQl309gtKi1kn8JhQ.png" /><figcaption>ELEVATE Interface and Request Flow</figcaption></figure><h3>What Makes ELEVATE Special?</h3><ul><li><strong>Automation-Driven and Serverless Architecture </strong><br>ELEVATE uses smart automation and a serverless design, which means it’s efficient and cost-effective. It fits right into our existing AWS setup without extra hassle. Best of all, the monthly cost is less than $500, a significant saving compared to ready-made SaaS platforms. This means we get top-notch security and flexibility without spend high cost.</li><li><strong>Easy Integration</strong><br>ELEVATE works with our current identity and access systems (Azure AD &amp; AWS Identity Center), so there’s no need to start from scratch.</li><li><strong>Clear Approval Workflow</strong><br>The approval process is visual and straightforward, making it easy for managers to review and approve requests quickly.</li></ul><h3>Real Impact</h3><p>Rolling out ELEVATE to all our production environment accounts over 50 in just one month was a big milestone for our team. The results have been impressive:</p><ul><li><strong>Fewer security incidents, w</strong>ith permanent access removed, there are fewer chances for mistakes or unauthorized actions.</li><li><strong>Just-in-Time troubleshooting</strong> because team members can get the access they need, when they need it.</li><li><strong>Better compliance</strong> with security policies, since everything is tracked and auditable.</li></ul><p>For example, when a developer needed to fix an urgent issue in production, they requested access through ELEVATE, got special approval in minutes, and their access was automatically removed after the job was done. No more waiting, no more risk of forgotten permissions.</p><h3>Lessons Learned &amp; Future Enhancement</h3><p>Through our journey with ELEVATE, we’ve learned a few important things:</p><ul><li><strong>Keep it simple</strong>: The easier it is to use, the more people will follow the process.</li><li><strong>Automate everything you can</strong>: Automatic removal and logging save time and reduce errors.</li><li><strong>Review and improve</strong>: Regularly check the logs and workflows to spot areas for improvement.</li></ul><p>Looking ahead, our next plan is to integrate AI to compare Document MOPs (Methods of Procedure) with real activity logs. If there’s any discrepancy between what was planned and what actually happened, the system will <strong>automatically</strong> <strong>alert</strong> our <strong>Security Team</strong> or <strong>Productivity Team</strong>. This will help us catch issues even faster and keep our production environment even safer and well documented.</p><h3>Conclusion</h3><p>ELEVATE has helped us secure our production environment without slowing down our team. By giving just-in-time access, we keep things safe, fast, and under control. If your organization struggles with managing access, consider a temporary access system like <strong>ELEVATE</strong>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7cb7b68cac55" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/our-organization-secures-production-environment-with-elevate-temporary-elevated-access-management-7cb7b68cac55">Our Organization Secures Production Environment with ELEVATE Temporary Elevated Access Management</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cloud Trial and Exploration Framework: Accelerating Innovation at Telkomsel]]></title>
            <link>https://medium.com/life-at-telkomsel/cloud-trial-and-exploration-framework-accelerating-innovation-at-telkomsel-94f5a245b3a2?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/94f5a245b3a2</guid>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[innovation]]></category>
            <dc:creator><![CDATA[Putu Adhika]]></dc:creator>
            <pubDate>Wed, 10 Sep 2025 08:18:32 GMT</pubDate>
            <atom:updated>2025-09-10T07:54:56.215Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/570/1*KZVTr9ZTcgUZaKyaRt73tQ.jpeg" /></figure><p>At Telkomsel, innovation has always been one of the strongest pillars that keeps us moving forward in an increasingly competitive digital era. To continue leading the market, it is not enough to rely only on existing solutions. We need to continuously find business opportunities by exploring new technologies, validate new approaches, and test bold ideas that may become the foundation for future growth. Cloud computing has become a powerful enabler in this journey, offering flexibility, scalability, and advanced services that allow teams to innovate without heavy upfront investments. This is more than just technology enablement. It is a cultural shift that empowers every team to innovate.</p><p>Recognising this, our Cloud Center of Excellence (CCoE) has introduced the Cloud Trial and Exploration program, a structured initiative that provides sandbox environments to all directorates in Telkomsel. Through this program, teams can safely run Proof of Concepts (POCs) and test innovation use cases across the leading cloud platforms such as Amazon Web Services (AWS), Microsoft Azure, GCP, and Huawei Cloud. By providing these sandbox environment to all teams that needed, we foster a culture of innovation that spans across the entire organization, rather than being confined to a single function.</p><p>The sandbox is not just a playground. It is a place where ideas can be tested against reality. A new customer experience feature can be validated, a new machine learning model can be trained, or a next-generation digital services can be simulated before deciding whether it should scale into production. All of this is supported by a governance framework that makes sure every trial follows a clear flow, with well-defined entry points, assessments, implementation steps, and reporting. This approach helps us strike the right balance between agility and accountability.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mtLi7TRZgDW2oGoVBNdQ7g.png" /><figcaption>Cloud Trial and Exploration Framework</figcaption></figure><p>During the initialization and assessment phases, the user story or use case to be deployed in the sandbox is defined. The CCoE team then review the solution design to ensure it aligns with cloud architecture best practices and provides recommendations where necessary. Users are required to submit a form detailing the cloud services to be utilized, cost estimation, project timeline, and user access requirements.</p><p>In the implementation phase, the CCoE team prepares and provisions a dedicated sandbox environment. This environment is fully isolated from Telkomsel’s on-premise data centers and production accounts to maintain security and prevent any operational impact. This approach empowers team to innovate and iterate freely, while adhering to governance and compliance standards.</p><p>Once the setup is complete, users can begin their trial or POC within the sandbox. It provides a safe space for cross-functional teams to collaborate, learn cloud-native practices, and validate configurations before promoting them to production environments. Upon completion of the POC or trial, the CCoE team facilitates a structured review process. This includes the preparation of a comprehensive POC report that documents the outcomes, challenges, and lessons learned. The report also includes a set of recommendations that determine whether the solution is viable for production deployment.</p><p>If the POC is deemed successful, the next step involves a collaborative discussion between stakeholders to evaluate the feasibility of scaling the solution into a production-ready architecture. This includes considerations around performance, security, cost optimization, operational readiness, and alignment with Telkomsel’s cloud governance policies. The goal is to ensure that the solution can be reliably scaled and maintained in a live environment.</p><p>Since the program was rolled out in 2021, the adoption has been very encouraging. So far, Telkomsel has completed more than 100 cloud trials in AWS, Azure, and Huawei Cloud combined. Each of these trials has contributed valuable learnings, and several have gone beyond the POC stage to become fully deployed production solutions. This means that the cloud trial initiative is not only helping us test ideas, but also directly generating impact on our business operations and innovation driver.</p><p><strong>The Benefits of Cloud Trial</strong></p><p>What makes the Cloud Trial program especially impactful is the tangible business value it has already generated. The benefits can be seen in three major dimensions:</p><p>First, assessing new technologies. Cloud gives Telkomsel teams rapid access to emerging capabilities such as generative AI with Amazon Bedrock and Azure AI Foundry, privacy-focused collaboration through AWS Clean Rooms, and advanced machine learning pipelines powered by Amazon SageMaker AI. Instead of waiting months to procure and set up infrastructure, teams can now explore these cutting-edge technologies within days, even seconds!</p><p>Second, monetization. From the total cloud trials, several POC initiatives have successfully progressed to production, with some of them now serving as new revenue generators. Initiatives such as Telkomsel DFE, SSP, MEPRO, and DXP are concrete examples of how experimentation in the sandbox can translate into real business outcomes. These projects prove that innovation in the cloud is not only about exploration but also about unlocking new business models and revenue streams.</p><p>Third, innovation program support. The cloud trial also fuels broader innovation movements within Telkomsel. Some initiatives have been supported to drive innovation programs in Telkomsel like Polaris and InnovationHub. Participants of Polaris and InnovationHub can take advantage of this sandbox environment to easily develop their MVPs, leveraging the flexibility and convenience of the cloud.</p><p><strong>Building an Innovation Culture</strong></p><p>But beyond frameworks and numbers, the Cloud Trial initiative embodies something much bigger, Telkomsel’s innovation culture. By lowering the barrier for teams to experiment and by providing access to new technologies, the program fosters an environment where curiosity and creativity thrive.</p><p>This mindset is crucial for Telkomsel to sustain its market leadership. Innovation is not just about incremental improvements. It is about finding new alternatives, improving existing solutions, and exploring <em>blue ocean strategies</em> that can redefine our position in the market. With cloud as the backbone, we are ensuring that these explorations are both scalable and sustainable.</p><p>As more directorates in Telkomsel embrace the Cloud Trial and Exploration program, the opportunities for breakthrough innovation will continue to expand. With successful POCs already creating new revenue streams and supporting strategic programs, we see the initiative as a strategic engine for both technological advancement and business growth.</p><p>At Telkomsel, innovation doesn’t just happen. It is engineered, tested, and scaled through the cloud. And with the Cloud Trial and Exploration Framework, we are making sure that every idea gets the chance to prove its value.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=94f5a245b3a2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/cloud-trial-and-exploration-framework-accelerating-innovation-at-telkomsel-94f5a245b3a2">Cloud Trial and Exploration Framework: Accelerating Innovation at Telkomsel</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Understanding the True Cost of a Kubernetes Workload]]></title>
            <link>https://medium.com/life-at-telkomsel/understanding-the-true-cost-of-a-kubernetes-workload-3a81e2b9529b?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/3a81e2b9529b</guid>
            <category><![CDATA[api]]></category>
            <category><![CDATA[observability]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <dc:creator><![CDATA[Lutfi Ichsan Effendi]]></dc:creator>
            <pubDate>Tue, 05 Aug 2025 08:54:18 GMT</pubDate>
            <atom:updated>2025-08-23T00:46:15.623Z</atom:updated>
            <content:encoded><![CDATA[<h4>Trace individual microservice costs by combining Kubernetes metrics, APM, and CUR for granular spending insights</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XQpMa57dQM8dPv0IMiHyrg.png" /></figure><p>Kubernetes is a powerful orchestration platform, enabling rapid deployment, scalability, and management of containerized applications. But its abstraction comes with a hidden cost. While teams reap the benefits of automated infrastructure, they often lose visibility into how much each service, node, or API actually costs. Understanding the true cost of running Kubernetes workloads isn’t just a matter of keeping cloud bills in check, it’s foundational to making informed engineering and business decisions.</p><p>This article dives deep into how we can trace cloud spend from cluster level resources down to individual API calls, combining Kubernetes metrics, APM data and cloud billing reports (like AWS CUR). It’s a critical step toward maturing your FinOps practice and achieving sustainable, optimized cloud operations.</p><h4>1. Why Kubernetes Cost Visibility is Critical</h4><p>Kubernetes operates on a shared resource model. Multiple services, teams, and applications co-exist in the same cluster, consuming CPU, memory, storage and network bandwidth. But, unlike traditional VM or isolated instances, Kubernetes doesn’t naturally expose the cost per service or per API call.</p><p>Most organizations monitor their cloud spending at the cluster or account level, which makes it impossible to answer questions like :</p><ul><li>How much does our recommendation engine cost to run per day?</li><li>Which team services are overprovisioned?</li><li>What is the cost impact of a sudden spike in a single API usage?</li></ul><p>To solve this, we need a cost observability framework that breaks down usage and spending across layers :</p><blockquote>Cluster → Node → Service → DTO → API</blockquote><h4>2. Cost Breakdown in Kubernetes: From Cluster to API Level</h4><h4>2.1 Cluster-Level Costs</h4><p>At the top of the hierarchy are the raw compute resources :</p><ul><li>CPU, memory, persistent storage and inter AZ networking</li><li>Control plane (managed by provider) vs worker nodes (billed per usage)</li><li>Managed Kubernetes pricing (e.g. EKS charges $0.10/hour for control plane)</li></ul><p>This forms the baseline of your cloud bill, but tells you little about which workloads drive those costs.</p><h4>2.2 Node-Level Costs</h4><p>Nodes differ by :</p><ul><li><strong>Type </strong>: On Demand, Spot, Reserved, Graviton (for AWS)</li><li><strong>Provisioning strategy </strong>: Cluster Autoscaler vs Karpenter (or similar tools)</li><li><strong>Architecture </strong>: x86 vs ARM, affecting cost/performance ratio</li></ul><p>Overprovisioned nodes or inefficient autoscaling policies can lead to wasted spend that’s invisible unless properly tracked.</p><h4>2.3 Service-Level Costs</h4><p>Each microservice declares its own resource requests/limits. Meanwhile, in practice:</p><ul><li>Many services over-request resources “just to be safe”</li><li>Others under request, triggering noisy neighbor problems</li></ul><p>Also, service-to-service traffic often managed via a service mesh (e.g. Istio, Consul) consumes CPU, memory and networking overhead. These indirect costs often go unmonitored.</p><h4>2.4 DTO-Level Costs (Data Transfer Objects)</h4><p>Not all data movement is free :</p><ul><li>Internal traffic (east-west) can add latency and cost</li><li>External data egress (e.g. S3, internet bound) is expensive</li></ul><p>Ignoring DTO sizes and frequencies may result in large bills with no obvious culprit.</p><h4>2.5 API-Level Costs: Estimating Cost per Endpoint with Custom Attribution</h4><p>Every API endpoint in a microservice doesn’t just process data, it consumes compute, memory, storage and network resources. Some endpoints are simple reads from cache, others trigger multiple downstream calls (e.g. database writes, third-party API invocations or queue operations). These differences result with different resource consumption profiles and costs per API.</p><p>Yet, cloud providers and Kubernetes do not natively expose per endpoint cost visibility.</p><p><em>So, how can we approximate the cost of an individual API?</em></p><p>A practical way to calculate API level costs is to divide the total service cost by observed API call frequencies, using a custom attribution ratio. Here is how :</p><p><strong>Step 1: Calculate Total Cost of the Service</strong></p><p>Start by calculating the total cost for running a single service over a given time window (e.g. daily or hourly), combining data from :</p><ul><li><strong>Cloud billing (CUR) </strong>: Assign costs for compute, memory, and network to the service/Kubernetes deployment object (based on tags, labels, or container IDs)</li><li><strong>Prometheus/Kube-state-metrics </strong>: Track actual CPU/memory usage</li></ul><p>Let’s say :</p><blockquote><em>The payment-service costs </em><strong><em>$120/day</em></strong><em> in total resources.</em></blockquote><p><strong>Step 2 : Capture API Call Distribution via APM</strong></p><p>Use APM tools (like Datadog APM, Dynatrace, OpenTelemetry or Jaeger) to observe :</p><ul><li>All API endpoints</li><li>Total request count per endpoint in the same time window</li></ul><p>Example (daily request volume):</p><ul><li>/checkout: 10,000 calls</li><li>/refund: 2,000 calls</li><li>/status: 5,000 calls</li></ul><p>Total API calls: <strong>17,000</strong></p><p><strong>Step 3: Apply Proportional Cost Attribution</strong></p><p>Now calculate cost per API by proportional distribution :</p><pre>Cost per call = Total service cost / Total number of calls</pre><pre>/checkout → (10,000 / 17,000) × $120 = $70.59<br>/refund → (2,000 / 17,000) × $120 = $14.12<br>/status → (5,000 / 17,000) × $120 = $35.29</pre><p>You can go further to get cost per request :</p><pre>/checkout → $70.59 / 10,000 = $0.007 per call</pre><blockquote>This assumes a uniform cost per request, which isn’t always true.</blockquote><p>For a more accurate and insightful model, enrich this cost attribution with additional weights :</p><p><strong>Option A: Weighted by Resource Consumption</strong></p><p>Track resource usage per endpoint using APM integrated metrics (OpenTelemetry spans with CPU/mem usage as custom tags). Adjust the ratio :</p><pre>Weighted Cost = (Request count × Avg CPU usage × weight_factor) / Total weighted usage × Total cost</pre><p>This way, endpoints that invoke heavy database operations or downstream systems are attributed a higher share of the cost.</p><p><strong>Option B: Tie to Business Value (Cost per Outcome)</strong></p><p>Incorporate business metrics such as :</p><ul><li>Revenue per API call</li><li>Conversion rate</li><li>Customer segment (e.g., VIP vs. regular)</li></ul><p>Example :</p><p>If /checkout leads to $5 average revenue and costs $0.007/call → profit margin = <strong>$4.993</strong></p><p>If /refund costs $0.007/call but recovers $0 → it’s a pure expense, worth optimizing or rate limiting</p><h4>3. Insights &amp; Optimization Strategies</h4><p>Once the framework is in place, it opens up powerful FinOps opportunities :</p><ul><li><strong>Detect overprovisioning </strong>: Services requesting 500m CPU but only using 100m</li><li><strong>Right-size with autoscaling </strong>: Use HPA/VPA to adjust in real time</li><li><strong>Adopt Spot or Graviton</strong>: Identify workloads safe for cost optimized nodes</li><li><strong>Reduce data egress </strong>: Flag services pushing excessive outbound data</li><li><strong>FinOps benchmarking </strong>: Which endpoints generate the highest cost? Which generate the highest value?</li><li><strong>Engineering prioritization </strong>: Identify APIs to optimize first (e.g. caching or code efficiency)</li><li><strong>Business discussions </strong>: Show product managers or execs the cost of scaling a new feature endpoint</li></ul><p>These insights can be looped back into automation pipelines, enabling cost aware deployments and capacity planning.</p><h4>Real Time Kubernetes FinOps</h4><p>Cost observability in Kubernetes isn’t a nice to have, it’s essential for maintaining sustainable cloud operations at scale. While the abstraction of containers and microservices brings flexibility, it also obscures cost accountability.</p><p>By combining telemetry (Kubernetes metrics), tracing (APM), and billing data (CUR), organizations can achieve API level cost insights and make data driven decisions for optimization.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3a81e2b9529b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/understanding-the-true-cost-of-a-kubernetes-workload-3a81e2b9529b">Understanding the True Cost of a Kubernetes Workload</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Operational Excellence Unleashed: How Observability Drives Scalability, Cost Efficiency, and Rapid…]]></title>
            <link>https://medium.com/life-at-telkomsel/operational-excellence-unleashed-how-observability-drives-scalability-cost-efficiency-and-rapid-52e3cd6e17e7?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/52e3cd6e17e7</guid>
            <category><![CDATA[cost-efficiency]]></category>
            <category><![CDATA[operational-excellence]]></category>
            <category><![CDATA[scalability]]></category>
            <category><![CDATA[management]]></category>
            <dc:creator><![CDATA[Christian Mahardhika]]></dc:creator>
            <pubDate>Thu, 03 Jul 2025 09:16:55 GMT</pubDate>
            <atom:updated>2025-07-03T09:16:55.640Z</atom:updated>
            <content:encoded><![CDATA[<h3>Operational Excellence Unleashed: How Observability Drives Scalability, Cost Efficiency, and Rapid Repair</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DVI8VvUl-IkqLZAwa59I5w.png" /></figure><p>Are You Truly Ready for the Next Major Incident?</p><p>In today’s fast-paced digital world, traditional monitoring simply isn’t enough. Without robust observability engineering, your business could be exposed to unexpected outages, escalating costs, and prolonged downtime — risks that can erode customer trust and damage your reputation in moments.</p><p>The truth is, every missed anomaly or undetected issue can quickly escalate, threatening the stability of your entire production system. Modern observability engineering isn’t just a nice-to-have; it’s essential for maintaining control, making informed scalability decisions, reducing incident response times, and resolving bugs before they impact your users.</p><p>This in-depth analysis reveals why forward-thinking organizations are turning to observability engineering to stay ahead of problems, optimize operations, and ensure business continuity. Don’t wait for a crisis to reveal the gaps in your systems — discover how comprehensive data collection and intelligent analysis can safeguard your company’s future.</p><p>Keep reading — because the health of your business may depend on it.</p><h3>Observability Engineering: The Foundation of Modern Application Maintenance and Production Excellence</h3><p>Observability engineering is revolutionizing how digital organizations maintain their applications and production systems by shifting the focus from reactive firefighting to proactive problem prevention. Instead of waiting for alarms to trigger and scrambling to fix issues after they disrupt business operations, modern observability platforms empower teams to harness the vast amounts of data already being collected across their environments. By continuously analyzing this data, IT teams can detect subtle warning signs, identify patterns that precede incidents, and address potential problems before they impact users or trigger alarms. This proactive approach not only reduces downtime but also enables organizations to optimize performance, improve reliability, and stay ahead of issues that could threaten business continuity.</p><p>At the heart of observability are three core pillars: metrics, logs, and traces. <strong>Metrics</strong> provide quantitative measurements of system performance and health. For example, teams can monitor the response time of each API endpoint, track resource utilization such as CPU, memory, and disk usage across infrastructure, pods, and applications, and keep a close watch on uptime or service availability. These metrics offer real-time insights into how well the system is performing and where potential bottlenecks may arise. <strong>Logs</strong> capture detailed messages or stack traces about what is happening within the system, especially when errors occur. For instance, when a service fails, logs can provide the exact error message, the sequence of events leading up to the failure, and even the specific line of code involved — making troubleshooting and root cause analysis much more efficient. <strong>Traces</strong> deliver end-to-end visibility into request flows across complex, distributed systems. This includes tracking inter-service communications, following the flow of data through multiple microservices, and even drilling down to see how each function or method within the code is executed during a transaction. Traces are invaluable for understanding latency issues, identifying performance bottlenecks, and mapping service dependencies in real time.</p><p>The rise of microservices, cloud computing, and containerization has made IT environments more dynamic and interconnected than ever before. However, observability’s role extends beyond managing technical infrastructure complexity — it also addresses the challenges posed by rapidly evolving business models and processes. As business processes grow more complex and change quickly, these shifts can directly influence application performance and system behavior. Observability platforms help organizations monitor not only the technical stack but also how business process changes impact system health. Additionally, in today’s interconnected digital ecosystem, applications often integrate with external systems operated by partners or other companies. Observability enables teams to maintain optimal performance across these integrations, quickly detect and diagnose errors that may arise in partner systems, and proactively collaborate to resolve issues. This shared visibility and responsiveness help reduce service-level agreement (SLA) breaches, strengthen partnerships, and ensure smoother, more reliable end-to-end service delivery.</p><p>The business impact of adopting observability-driven maintenance strategies is profound. Organizations with mature observability practices experience significantly less downtime — up to 79% less — compared to those relying on legacy monitoring solutions. This translates directly into cost savings, improved customer satisfaction, and enhanced business continuity, all of which are critical in today’s highly competitive digital marketplace.</p><p>To fully realize these benefits, IT leaders must implement best practices for production monitoring. This includes defining clear objectives for system performance, establishing robust monitoring strategies, and setting key performance indicators such as response time, availability, error rates, and user satisfaction. Continuous analysis of these metrics enables data-driven decision-making, targeted system improvements, and a proactive approach to issue detection and resolution.</p><h3>Uptime Monitoring, Service Performance, and Scalability: The Observability Advantage</h3><p>Observability engineering is fundamentally changing how organizations approach <strong>uptime</strong> and <strong>availability management</strong>. Imagine running a busy airport: if you only react when a plane is delayed, chaos can quickly follow. Observability is like having a <strong>control tower</strong> that not only tracks every plane in real time but also <strong>predicts potential delays</strong> before they happen. Instead of waiting for alarms to go off after a system outage, observability platforms allow teams to watch the “traffic” of data, spot <strong>early warning signs</strong>, and <strong>prevent downtime</strong> before users even notice a problem. This <strong>proactive approach</strong> is what keeps digital businesses running <strong>smoothly and reliably</strong>.</p><p><strong>Service performance monitoring</strong> with observability is like monitoring the flow of cars through a city’s traffic lights. You don’t just want to know if there’s a traffic jam after it happens — you want to see how traffic is moving at every intersection, in real time. Observability platforms provide this kind of <strong>real-time visibility</strong> for applications, showing how quickly transactions are processed, whether <strong>resources</strong> like servers and databases are being used efficiently, and if users are experiencing <strong>slowdowns</strong>. With this information, organizations can set <strong>Service Level Objectives (SLOs)</strong> — like promising that 99.9% of users will get a response within two seconds — so everyone knows what “good performance” really means.</p><p>When it comes to <strong>measuring service performance</strong>, observability is like having a dashboard in your car that not only shows your speed but also warns you if the engine is overheating or if you’re running low on fuel. Organizations can track how often <strong>outages</strong> happen, how quickly problems are detected (<strong>mean time to detection</strong>), and how fast they’re fixed (<strong>mean time to resolution</strong>). Research shows that companies with <strong>full observability</strong> can fix major issues much faster — sometimes in less than 30 minutes — because they have all the right “gauges” and <strong>alerts</strong> in place to guide their actions.</p><p><strong>Application Performance Monitoring (APM)</strong> is another key benefit of observability. Think of it like a <strong>fitness tracker</strong> for your software: it continuously measures how well your applications are “running,” where they’re “out of breath,” and where they can be improved. With APM, teams can see exactly where an application is <strong>slowing down</strong> — maybe a database query is taking too long, or a server is overloaded — so they know where to focus their efforts for the biggest gains in <strong>speed and reliability</strong>.</p><p>Observability-driven APM also helps organizations <strong>save money</strong> and stay <strong>secure</strong>. It’s like having a smart thermostat that not only keeps your house comfortable but also tells you when you’re <strong>wasting energy</strong> or if there’s a possible <strong>break-in</strong>. By identifying where resources are being wasted or where <strong>unusual activity</strong> might signal a security issue, observability platforms help companies <strong>optimize costs</strong> and <strong>protect their systems</strong> from threats.</p><p>A major strength of observability is its ability to provide a <strong>“bird’s eye view”</strong> across all systems and users. Picture a theme park manager watching not just the rides, but also how crowds move, where lines form, and how weather affects attendance. Similarly, observability lets teams see how <strong>technical performance impacts user behavior and business outcomes</strong>, so they can <strong>prioritize improvements</strong> that matter most to customers and the bottom line.</p><p><strong>Scalability</strong> and <strong>infrastructure optimization</strong> are also made easier with observability. In cloud environments like Kubernetes, it’s like having a <strong>smart parking lot</strong> that automatically opens new spaces when more cars arrive and closes them when traffic slows down. Observability tools track <strong>usage patterns</strong> and help teams automatically scale resources up or down, ensuring <strong>efficiency</strong> and <strong>controlling costs</strong>. For databases, observability is like monitoring the checkout lines at a supermarket: if lines get too long, you know to open more registers before customers get frustrated.</p><p>Finally, <strong>comprehensive infrastructure monitoring</strong> is like having security cameras and sensors throughout a building, tracking everything from the front door to the basement. Modern observability platforms <strong>visualize network traffic</strong>, detect <strong>unusual patterns</strong>, and help teams <strong>spot potential issues</strong> before they become big problems. This <strong>layered approach</strong> — from the deepest technical details to high-level business metrics — ensures that organizations can <strong>scale confidently</strong> and deliver <strong>top-notch performance</strong> at every level.</p><h3>Rapid Bug Resolution Without Error Reproduction and MTTR Reduction: The Observability Breakthrough</h3><p>One of the most <strong>transformative benefits</strong> of observability engineering is the ability to achieve <strong>rapid bug resolution</strong> and dramatic <strong>Mean Time to Repair (MTTR) reduction</strong> — all without the time-consuming need to <strong>reproduce errors</strong>. Traditionally, debugging software was like trying to solve a mystery with only a few clues — developers would spend hours or even days attempting to recreate the exact conditions that triggered a failure. With advanced observability, it’s as if you have a <strong>full surveillance system</strong> capturing every detail as incidents unfold, allowing teams to pinpoint the root cause almost instantly. Organizations leveraging <strong>comprehensive observability solutions</strong> have reported up to <strong>95% improvement in resolution times</strong>, shrinking incident response from hours to just minutes.</p><p>This breakthrough is possible because observability enables <strong>real-time data collection and analysis</strong>, giving teams immediate insight into what’s going wrong. Imagine a car dashboard that not only flashes a warning light but also displays which part is failing and suggests a fix before you break down. With this <strong>rapid response capability</strong>, the impact of incidents on business operations is minimized, and <strong>business continuity</strong> is maintained.</p><p>Modern observability platforms provide <strong>comprehensive monitoring and visualization</strong> not just for technical systems, but for <strong>business processes</strong> as well. Think of it as a smart building management system that watches every elevator, door, and HVAC unit, alerting you to problems before tenants complain. <strong>Predictive analytics</strong> continuously ingest data like CPU, memory, network metrics, and application logs, using <strong>time-series forecasting</strong> and <strong>anomaly detection</strong> to flag potential faults with over <strong>96% accuracy</strong>.</p><p>Taking this a step further, <strong>self-healing infrastructure</strong> — enabled by observability — acts like a car that can automatically pull over and fix a flat tire without the driver lifting a finger. By combining <strong>predictive analytics</strong>, <strong>reinforcement learning</strong>, and <strong>rule-based automation</strong>, organizations have achieved <strong>85% MTTR reduction</strong>, with recovery reliability exceeding <strong>95%</strong> and fault tolerance above <strong>91%</strong>. The integration of <strong>artificial intelligence (AI)</strong> and <strong>machine learning (ML)</strong> in observability-driven incident management means systems can learn from operational data and <strong>continuously improve</strong> their own processes, much like a self-driving car that gets better with every mile.</p><p>Observability doesn’t just make systems faster to fix — it makes them <strong>smarter at preventing issues</strong> altogether. <strong>Automated incident management</strong> uses machine learning to analyze mountains of operational data, identify patterns, and predict potential issues before they occur. This shifts teams from a <strong>reactive firefighting mode</strong> to a <strong>proactive prevention mindset</strong>, reducing the number and severity of incidents.</p><p>The heart of rapid bug resolution without error reproduction lies in <strong>comprehensive data collection</strong> and <strong>distributed tracing</strong>. Observability platforms capture the full context of every transaction as it happens — like having a flight recorder on an airplane. When something goes wrong, you can replay the entire sequence of events, pinpointing exactly where and why the failure occurred — no guesswork, no need to recreate the scenario.</p><p><strong>Context-aware debugging</strong> becomes possible with rich trace data that shows not just what failed, but how different services interacted and what led up to the issue. Imagine a detective with access to every security camera, phone log, and witness statement — solving the case becomes much easier. Automated test frameworks can even inject traces and logs during testing, providing real user-like telemetry that accelerates <strong>root cause identification</strong>.</p><p>Finally, the <strong>correlation of metrics, logs, and traces</strong> allows teams to conduct thorough incident retrospectives. It’s like having a complete play-by-play for every game, enabling coaches to spot weaknesses, adjust strategies, and prevent future losses. This data-driven approach ensures that every incident is a learning opportunity, leading to <strong>continuous improvement</strong> in both systems and processes.</p><h3>Implementation Strategies and Operational Excellence Framework</h3><p>Implementing <strong>enterprise observability</strong> is not just about plugging in new tools — it’s about orchestrating a <strong>comprehensive transformation</strong> across both technology and organizational culture. Imagine a city upgrading from scattered neighborhood watch groups to a unified, citywide surveillance and emergency response system. Organizations that reach <strong>observability maturity</strong> are more than twice as likely to use <strong>centralized observability approaches</strong> rather than leaving each team to fend for itself. This centralized model is like having a central control room for the city, allowing for advanced techniques such as <strong>distributed tracing</strong> — where every “street” and “intersection” in your IT landscape is monitored and connected for rapid, coordinated response (<a href="https://grafana.com/blog/2025/03/25/observability-survey-takeaways/">Grafana, 2025</a>).</p><p>But <strong>successful observability adoption</strong> goes far beyond technology. It’s a <strong>cultural shift</strong>: teams must move from simply reacting to problems to proactively anticipating and resolving them. Think of it like a hospital that doesn’t just treat patients when they arrive, but uses health data to predict and prevent outbreaks. Companies that embrace observability can <strong>optimize performance</strong> and <strong>enhance customer experiences</strong> in ways that drive real business value, shifting from reactive “firefighting” to <strong>proactive operational management</strong>.</p><p>The <strong>maturity</strong> of observability practices can be measured by how organizations collect and use telemetry. Most companies start by focusing on <strong>metrics</strong> (like a car’s speedometer) and <strong>logs</strong> (like a black box recording every event). As they mature, they adopt <strong>distributed tracing</strong> — which is like having a GPS history for every journey through your system. This <strong>three-pillar approach</strong> (metrics, logs, traces) shows an organization is ready for <strong>advanced operational excellence</strong>, with a “map” detailed enough to navigate any challenge (<a href="https://grafana.com/blog/2025/03/25/observability-survey-takeaways/">Grafana, 2025</a>).</p><p>For observability to truly deliver business value, it must be <strong>strategically aligned</strong> with business objectives. Imagine a chef who not only checks the temperature of the oven but also monitors customer satisfaction and ingredient freshness in real time. Organizations must integrate observability across <strong>diverse data sources</strong> — from time-series databases and log management systems to business KPIs and service mesh telemetry. Smart strategies like <strong>differential updates</strong>, <strong>importance-based refresh rates</strong>, and <strong>contextual caching</strong> help manage the complexity, ensuring that the “kitchen” runs smoothly without being overwhelmed by too much data.</p><p>A key value of observability is its role in building <strong>impactful dashboards and KPIs</strong>. Think of a pilot’s cockpit, where every dial and gauge reflects the real-time health of the aircraft. <strong>Data observability</strong> ensures that business dashboards are not just pretty charts, but <strong>accurate, reliable, and timely</strong> instruments that guide decision-making. This allows leaders to uncover hidden insights, spot problems before they escalate, and act with confidence (<a href="https://sixthsense.rakuten.com/blog/The-Role-of-Data-Observability-in-Building-Impactful-Dashboards-and-KPIs">Rakuten, 2024</a>).</p><p>Looking ahead, the future of observability lies in <strong>integration with artificial intelligence and advanced analytics</strong>. Imagine a smart city where AI analyzes traffic, weather, and social media to optimize everything from emergency response to energy use. For digital businesses, this means <strong>competitive advantages</strong> through better customer insights, operational efficiency, and personalized engagement — directly translating to <strong>higher retention and revenue growth</strong>.</p><p>However, achieving this vision comes with <strong>significant challenges</strong>, especially when it comes to <strong>storage capacity limitations</strong>. Modern observability platforms generate a <strong>massive volume of telemetry data</strong> — logs, metrics, and traces — that can quickly overwhelm traditional storage systems as applications and infrastructure scale. For example, large enterprises may generate <strong>terabytes of observability data every day</strong>, with some organizations storing over 10TB daily just for monitoring and analytics purposes (<a href="https://blog.quest.com/common-observability-challenges-and-steps-to-overcome-them/">Quest, 2023</a>). This relentless data growth leads to <strong>increased storage requirements</strong>, higher costs, and potential <strong>performance bottlenecks</strong> that can hinder timely insights and slow down troubleshooting. The <strong>sheer data volume</strong> can cause <strong>latency</strong> and <strong>reduced responsiveness</strong>, making it harder for teams to extract actionable information when it matters most. To address these issues, organizations must adopt <strong>robust data retention policies</strong>, such as archiving or deleting redundant information and storing only high-value data for extended periods (<a href="https://www.kloudfuse.com/blog/how-to-architect-observability-storage-with-ebs-without-hitting-aws-limits">Kloudfuse, 2024</a>). <strong>Tiered storage architectures</strong> are often used, where recent, high-resolution data is kept in fast-access databases, while older or less critical data is moved to lower-cost, long-term storage like Amazon S3 or Azure Blob Storage. This is similar to a library that keeps the most-read books on the main shelves and archives rarely accessed ones in the basement — ensuring quick access to what’s important while controlling costs. Another best practice is to <strong>separate observability data storage</strong> from main transactional and analytic databases, allowing each to be optimized for its unique workload and access patterns (<a href="https://dqops.com/data-observability-benefits/">DQOps, 2024</a>). This separation is like having dedicated warehouses for perishable goods, electronics, and archival records — each with its own climate control and security, ensuring efficiency and reliability. Additionally, <strong>data filtering and sampling</strong>, <strong>aggregation</strong>, and <strong>intelligent retention strategies</strong> help reduce the volume of stored data without sacrificing critical visibility. Despite advances in cloud storage, <strong>cost management remains a major concern</strong>: cloud-native solutions may offer “infinite” capacity, but budgets are not infinite, and storage costs can escalate rapidly if not carefully managed (<a href="https://www.observeinc.com/blog/lost-logs-retention-vs-cost/">Observe, 2024</a>). Furthermore, long-term log retention is crucial for <strong>anomaly detection</strong>, <strong>root cause analysis</strong>, and <strong>trend analysis</strong> — but the longer data is kept online, the higher the storage bill. Organizations must balance the need for <strong>comprehensive historical data</strong> with the realities of <strong>storage limits and cost efficiency</strong>, often leveraging <strong>observability pipelines</strong> to route, filter, and shape data before it is ingested and stored (<a href="https://www.exabeam.com/blog/siem-trends/overcoming-data-limitations-with-observability-pipelines/">Exabeam, 2024</a>). Finally, as observability systems scale, <strong>performance issues</strong> can arise in querying and retrieving large datasets, especially during peak loads or when trying to analyze historical trends (<a href="https://lumigo.io/what-is-observability-concepts-use-cases-and-technologies/">Lumigo, 2025</a>). This makes it essential to choose storage solutions and architectures that are <strong>scalable, performant, and cost-effective</strong>, and to continuously monitor and adjust storage strategies as the business and its systems evolve. Without these best practices, organizations risk losing the very visibility and responsiveness that observability is meant to provide.</p><p>Through the <strong>comprehensive implementation</strong> of observability engineering — spanning application maintenance, scalability, cost optimization, MTTR reduction, and rapid bug resolution — organizations unlock new levels of <strong>operational excellence</strong> and <strong>business performance</strong>. Observability becomes the backbone of a modern, proactive, data-driven enterprise, enabling sustainable competitive advantages in an increasingly <strong>complex digital world</strong>.</p><h3>Summary: Observability as the Engine of Operational Excellence</h3><p><strong>Observability engineering</strong> has become a cornerstone of <strong>operational excellence</strong> in modern digital organizations, moving teams from reactive monitoring to proactive, data-driven management. By leveraging the triad of <strong>metrics, logs, and traces</strong>, observability platforms provide real-time visibility into system health, enabling early detection of anomalies, rapid root cause analysis, and prevention of outages before they impact users. This approach not only supports technical reliability but also aligns IT operations with business goals, ensuring uptime, optimizing resource usage, and supporting scalability in complex, interconnected environments.</p><p>Organizations that mature in their observability practices — through centralized strategies, cultural shifts, and integration with business KPIs — consistently report significant reductions in downtime and <strong>Mean Time to Repair (MTTR)</strong>, with some achieving up to 95% improvement in incident resolution speed. Observability also enables <strong>cost efficiency</strong> by supporting intelligent scaling, resource optimization, and targeted automation, while facilitating compliance with service-level agreements (SLAs) and enhancing customer trust.</p><p>Key to this transformation is the ability to <strong>anticipate and resolve issues proactively</strong>, rather than waiting for alarms or customer complaints. Observability empowers organizations to maintain high performance even as business models, integrations, and infrastructure rapidly evolve, and it supports continuous improvement by providing actionable insights for both technical and business stakeholders.</p><p>However, the journey to observability-driven excellence is not without challenges. The explosion of telemetry data demands robust storage strategies, cost management, and architectural best practices — such as separating observability data from transactional systems and using tiered storage — to ensure scalability and sustainability. Addressing these challenges is essential to fully realize the competitive advantages that observability can deliver.</p><p>In sum, <strong>observability is not just a technical upgrade, but a strategic imperative</strong> for businesses seeking agility, resilience, and leadership in the digital era. It underpins the ability to scale, control costs, and recover rapidly — making it a foundational pillar of operational excellence.</p><p><strong>support us/ buy me a coffee?<br></strong><a href="https://sociabuzz.com/christianmahardhika/tribe">https://sociabuzz.com/christianmahardhika/tribe</a></p><h3>References</h3><ol><li><a href="https://www.techtarget.com/searchitoperations/tip/Top-observability-tools">Top observability tools</a></li><li><a href="https://em360tech.com/tech-articles/observability-secret-weapon-proactive-incident-management">Observability: The secret weapon for proactive incident management</a></li><li><a href="https://coralogix.com/guides/observability/">Observability Guide</a></li><li><a href="https://www.splunk.com/en_us/blog/devops/using-observability-to-reduce-mttr.html">Using observability to reduce MTTR</a></li><li><a href="https://en.blog.mrsuricate.com/monitoring-de-production-logiciel-guide-ultime">Ultimate guide to production monitoring</a></li><li><a href="https://www.moesif.com/blog/technical/api-development/Top-Product-Observability-Tools-to-Boost-Business-Efficiency/">Top Product Observability Tools to Boost Business Efficiency</a></li><li><a href="https://www.ibm.com/think/topics/service-level-objective">Service Level Objective (IBM)</a></li><li><a href="https://newrelic.com/resources/report/observability-forecast/2023/state-of-observability/service-level-metrics">State of Observability: Service Level Metrics (New Relic)</a></li><li><a href="https://www.dynatrace.com/news/blog/what-is-apm/">What is APM? (Dynatrace)</a></li><li><a href="https://milvus.io/ai-quick-reference/how-does-observability-improve-database-scalability">How does observability improve database scalability? (Milvus)</a></li><li><a href="https://milvus.io/ai-quick-reference/how-does-observability-improve-database-scalability">How does observability improve database scalability? (Milvus)</a></li><li><a href="https://uptrace.dev/tools/top-observability-tools">Top Observability Tools (Uptrace)</a></li><li><a href="https://ritha.eu/journals/JORIT/issues/6/articles/4">JORIT: Infrastructure Observability Analytics</a></li><li><a href="https://www.alibabacloud.com/help/en/ack/ack-managed-and-ack-dedicated/user-guide/observability-1/">Observability Architecture in Kubernetes (Alibaba Cloud)</a></li><li><a href="https://em360tech.com/tech-articles/observability-secret-weapon-proactive-incident-management">Observability: The secret weapon for proactive incident management</a></li><li><a href="https://coralogix.com/guides/observability/">Observability Guide</a></li><li><a href="https://www.splunk.com/en_us/blog/devops/using-observability-to-reduce-mttr.html">Using observability to reduce MTTR</a></li><li><a href="https://jisem-journal.com/index.php/journal/article/view/9888">Predictive analytics and self-healing infrastructure</a></li><li><a href="https://www.mdpi.com/2076-3417/15/2/854">AI and ML in observability-driven incident management</a></li><li><a href="https://blogs.manageengine.com/opmanager-plus/2023/11/24/the-role-of-observability-in-incident-response.html">The role of observability in incident response</a></li><li><a href="https://ltu.diva-portal.org/smash/get/diva2:1964604/FULLTEXT01.pdf">Distributed tracing for error resolution</a></li><li><a href="https://dzone.com/articles/debugging-observability-first-systems">Debugging observability-first systems</a></li><li><a href="https://grafana.com/blog/2025/03/25/observability-survey-takeaways/">Observability survey takeaways (Grafana)</a></li><li><a href="https://sixthsense.rakuten.com/blog/The-Role-of-Data-Observability-in-Building-Impactful-Dashboards-and-KPIs">The Role of Data Observability in Building Impactful Dashboards and KPIs (Rakuten)</a></li><li><a href="https://blog.quest.com/common-observability-challenges-and-steps-to-overcome-them/">Common observability challenges and steps to overcome them (Quest)</a></li><li><a href="https://www.kloudfuse.com/blog/how-to-architect-observability-storage-with-ebs-without-hitting-aws-limits">How to Architect Observability Storage with EBS Without Hitting AWS Limits (Kloudfuse)</a></li><li><a href="https://dqops.com/data-observability-benefits/">Data Observability Benefits — Examples and Best Practices (DQOps)</a></li><li><a href="https://www.observeinc.com/blog/lost-logs-retention-vs-cost/">Lost Logs: Log Retention vs Observability Cost (Observe)</a></li><li><a href="https://www.exabeam.com/blog/siem-trends/overcoming-data-limitations-with-observability-pipelines/">Overcoming Data Limitations With Observability Pipelines (Exabeam)</a></li><li><a href="https://lumigo.io/what-is-observability-concepts-use-cases-and-technologies/">Observability in 2025: How It Works, Challenges and Best … (Lumigo)</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=52e3cd6e17e7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/operational-excellence-unleashed-how-observability-drives-scalability-cost-efficiency-and-rapid-52e3cd6e17e7">Operational Excellence Unleashed: How Observability Drives Scalability, Cost Efficiency, and Rapid…</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Code Smell — Dispensable]]></title>
            <link>https://medium.com/life-at-telkomsel/code-smell-dispensable-3e2ab499d0f1?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/3e2ab499d0f1</guid>
            <category><![CDATA[tips]]></category>
            <category><![CDATA[dispensable]]></category>
            <category><![CDATA[performance]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[code-smells]]></category>
            <dc:creator><![CDATA[Muhammad Hanif]]></dc:creator>
            <pubDate>Wed, 25 Jun 2025 01:50:39 GMT</pubDate>
            <atom:updated>2025-06-25T01:50:39.053Z</atom:updated>
            <content:encoded><![CDATA[<h3>Code Smells — Dispensables</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/679/1*Auu-_pQiwDgVRWnCKNVGGQ.jpeg" /><figcaption>Code smells everywhere…</figcaption></figure><p>When you hear “code smell”, what are your first thoughts? Mostly say it is useless codes that still run into program. Yeah, that’s one of the code smell, <strong>Dispensables</strong>.</p><blockquote><strong>Something pointless and unneeded whose absence would make the code cleaner, more efficient and easier to understand.</strong></blockquote><p>A dispensable code smell happens when unnecessary code accumulates. This leads to codebase clutter, which can slow down applications and negatively impact performance.</p><h3>The type of Dispensable</h3><h3><strong>1. Comments</strong></h3><blockquote>“<strong>The best comment is a good name for a method or class.”</strong></blockquote><p>The name of a method or attribute should be self-explanatory without needing comments. The same applies to the logic flow within a function. <strong>If the logic flow needs explanation, the code can be considered part of the Long Method smell and should be extracted</strong>. <strong>Comments may be used for documentation purposes</strong>.</p><p>For example, in the Java programming language, Javadoc using /** */ is used for documentation. The use of // for single-line or /* */ for multi-line comments should be avoided.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5DKU9eI-lm_zne_S3QlK2Q.png" /><figcaption>Sample of Comments</figcaption></figure><p><strong>Treatment for Too Many Comments: Extract Method and Rename Method<br></strong>From the code sample, we can extract methods such as creating new methods like printHeader(), printMenu(), and scanInput().</p><p>The built-in printMenu function can also be confusing for developers because it contains both printing and scanning logic. It might be better to rename it to printMenuAndGetInput().</p><h3><strong>2. Duplicate Code</strong></h3><blockquote>“<strong>A Duplicate Code smell represents code in multiple places that is the same or very similar”</strong></blockquote><p>Copy-paste, ctrl+c, ctrl+v, copy, tweak a bit :D — as long as it works, right? Do you do this often?</p><p>If you find yourself frequently copying and pasting code, remember the “Rule of Three.” <strong>When you’ve copied something for the third time, it’s a red flag — time to refactor and eliminate duplicate code</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/706/1*B8OOFksc-UVXL8zE1sKHPw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/1*ojhtIsIqftbJYgbFrgRfrw.png" /><figcaption>Left: sample, Right: treatment</figcaption></figure><h3><strong>3. Lazy Class</strong></h3><blockquote>“<strong>Lazy Class smells represent classes that are under-used”</strong></blockquote><p>Splitting a class is generally good, especially when aiming for the Single Responsibility Principle (SRP). However, <strong>having classes that are too small can also be problematic, as it increases the number of classes to read and makes it harder for programmers to understand the code</strong>.</p><p>A <em>lazy class</em> is a condition where a class has very minimal functionality — at the extreme, it might contain only one function. Lazy classes can occur as a result of refactoring, where a class that initially had many responsibilities ends up being split too far, resulting in a class with very limited functionality.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YqhJjUpz8oSRVje0luFmNA.png" /><figcaption>Sample of Lazy Class</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F4HbWFHhlWU_40sYaCnR6Q.png" /><figcaption>Treatment of Lazy Class</figcaption></figure><h3><strong>4. Data Class</strong></h3><blockquote>“<strong>A Data Class smell occurs when a class does not implement enough functionality itself to justify it being a class”</strong></blockquote><p>A data class refers to a class that only contains setters and getters. This class does not have any additional functions that can independently operate on the data within it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/842/1*rcwSM2Qb2nYJFLQgdGbqQQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/972/1*sMZOXWg-dPtyk4aEq9yqZg.png" /><figcaption>Left: sample, Right: treatment</figcaption></figure><h3><strong>5. Dead Code</strong></h3><blockquote>“<strong>A Dead Code smell occurs when code is not used, be it a variable, method, class, or parameter. This code smell is often created when code has become obsolete, and so is not instantiated, called, initialised, etc”</strong></blockquote><p>Our condolences for the code that was written but never used — it’s scary when developers aren’t careful and fail to realize the existence of dead code. Luckily, linters can sometimes help by highlighting dead code in grey, indicating that it’s unused.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*-qjrwNEbNY6p85sHvl6wrA.png" /><figcaption>Sample of Dead Code</figcaption></figure><p>Another example would be a requirement related to discounts: a 10% discount is given if the price is less than 10,000 and 20% if it’s more.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xcf6hUMg_14Ae43ggNZNjw.png" /><figcaption>Another sample of Dead Code</figcaption></figure><p>Suddenly, the requirement changes to a flat 15% discount for all, and somehow the programmer forgets to remove the previous logic.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*47NuDFCV_MDCrrfQQw2bzA.png" /><figcaption>Another sample of Dead Code of new changes</figcaption></figure><p>The discountPrice variable in the first condition will always be overwritten by the condition below it.</p><h3><strong>6. Speculative Generality</strong></h3><blockquote>“<strong>Speculative Generality is where code is written with so much caution for possible future changes, that the code becomes unnecessarily complex and harder to read”</strong></blockquote><p>Sometimes, a programmer tends to create classes, methods, fields, or parameters that are already prepared even though they haven’t been used yet. The programmer speculates that these functions might be needed someday. However, as the product evolves, those speculations often prove to be unnecessary. This leads to problems because we end up writing code that’s harder to read for something that isn’t even needed.</p><p><strong>Sample Case:</strong><br>Suppose there’s a requirement from the client <strong>stating that the price can have a currency of either IDR or USD</strong>.<br>The programmer speculates that IDR and USD are traditional currencies, and <strong>in the future, digital currencies like Bitcoin might be added</strong>. Based on this speculation, <strong>the programmer creates a class hierarchy</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*2Mn4Ctwnn3SxeI_PYU3ifw.png" /><figcaption>Overview of Speculative Generality</figcaption></figure><p>Code smell is something we must avoid when writing code, especially those categorized as “Dispensable.” <strong>The easiest way to prevent it is by regularly doing pair programming or code reviews before merging code with the main codebase</strong>.</p><p>In the AI era, we can also install more advanced linters in our IDEs, such as Copilot, Cursor or Windsurf, which can help programmers minimize code smells. The more consistently we practice these habits, the more they become second nature — resulting in cleaner code and improved program performance.</p><p><a href="https://refactoring.guru/refactoring/smells/dispensables">Dispensables</a></p><p>My other stories:</p><ul><li><a href="https://medium.com/life-at-telkomsel/aws-cloud-quest-gamification-for-training-and-certification-bc9d7ec858e8">AWS Cloud Quest: Gamification for Training and Certification</a></li><li><a href="https://hanifmhd.medium.com/how-secure-is-your-mobile-app-part-1-deb067c0fa4b">How Secure is Your Mobile App? — Part 1</a></li><li><a href="https://hanifmhd.medium.com/ios-development-for-beginner-swift-storyboard-part-1-cef1c09e199b">iOS Development for Beginner (Swift/Storyboard) — Part 1</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e2ab499d0f1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/code-smell-dispensable-3e2ab499d0f1">Code Smell — Dispensable</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Self-Healing Automation on Kubernetes Reduces Downtime]]></title>
            <link>https://medium.com/life-at-telkomsel/how-self-healing-automation-on-kubernetes-reduces-downtime-8fd0d163ab0a?source=rss----b17e01250ce9---4</link>
            <guid isPermaLink="false">https://medium.com/p/8fd0d163ab0a</guid>
            <category><![CDATA[serverless-architecture]]></category>
            <category><![CDATA[kubernetes-automation]]></category>
            <category><![CDATA[self-healing-systems]]></category>
            <category><![CDATA[incident-response]]></category>
            <dc:creator><![CDATA[Dwiki Kurnia]]></dc:creator>
            <pubDate>Mon, 16 Jun 2025 09:17:50 GMT</pubDate>
            <atom:updated>2025-06-16T09:17:50.130Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u5-E-pEEMtS82mux05op5A.jpeg" /></figure><p>In modern cloud infrastructure, failures are inevitable, but how they are handled defines system reliability. Kubernetes offers built-in self-healing mechanisms like pod restarts and rescheduling, but some issues require external intervention to restore services efficiently. This is where self-healing automation plays a crucial role in minimizing downtime and maintaining system stability.</p><p>At Telkomsel, we implemented a <strong>self-healing automation system</strong> that seamlessly integrates with our <strong>Kubernetes clusters</strong>, enabling precise <strong>remediation actions</strong> based on <strong>specific Kubernetes alerts</strong>.</p><p>Even with Kubernetes’ native capabilities like:</p><ul><li>Pod rescheduling</li><li>CrashLoopBackOff restarts</li><li>Liveness and readiness probes</li></ul><p>there are still gaps that can cause delays in remediation, for example Application-level anomalies (e.g., stuck business logic, memory leaks). These issues often require human intervention or external automation to resolve them effectively.</p><p>To solve this, we built an event-driven self-healing system using a combination of:</p><ul><li>Amazon EKS / Kubernetes</li><li>AWS Lambda</li><li>Amazon EventBridge</li><li>Amazon RDS</li><li>Amazon S3</li><li>Amazon SES</li><li>Amazon Bedrock Nova Pro</li><li>Telegram Bot API</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wz8jaRA1iO9DpJciGFBNBA.png" /><figcaption>Low Level Architecture &amp; Process</figcaption></figure><p><strong>How It Works (Step-by-Step)</strong></p><p><strong>Alert Triggered in Kubernetes:</strong></p><ul><li>Custom probe detects an anomaly (e.g., pod return error Socket Timed Out) and collect the data into Centralized DB.</li></ul><p><strong>Lambda Function Executes Remediation Logic:</strong></p><ul><li>The alert is converted into a custom event, triggering an AWS Lambda function.</li><li>AWS Lambda will send an alert to Telegram bot and sends real-time notifications to our on-call engineers.</li><li>Engineers can also respond with actions (e.g., “Approve Restart”, “Ignore”, “Escalate”)</li><li>Based on the selected response, AWS Lambda communicates with the Kubernetes API server to restart the specified pods.</li></ul><p><strong>Multi-Agent Follow-Up:</strong></p><ul><li>Agents may be scheduled to verify if remediation was successful, or retry if needed.</li><li>A summary notification is sent to all engineers after verification.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NC1Y4xAiLfAk0ZV__BtKHg.png" /><figcaption>Alert Notification &amp; Deletion Request</figcaption></figure><p><strong>Results and Benefits</strong></p><ul><li>Reduced Downtime: Faster resolution of Kubernetes-level issues (minutes to seconds), Possibility improve up to 83% faster restoration time.</li><li>Higher Reliability: Automation handles known issues predictably</li><li>Cost Savings Opportunity: Reduced manpower involvement and MTTR (Mean Time to Recovery)</li></ul><p><strong>Conclusion</strong></p><p>Kubernetes gives you the foundation for resilience, but coupling it with event-driven automation supercharges your ability to self-heal. At Telkomsel, this system has allowed us to move from reactive to proactive operations, reducing downtime while keeping engineers in the loop only when truly necessary.</p><p>If your Kubernetes setup still relies heavily on manual incident resolution, it’s time to rethink your automation strategy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8fd0d163ab0a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/life-at-telkomsel/how-self-healing-automation-on-kubernetes-reduces-downtime-8fd0d163ab0a">How Self-Healing Automation on Kubernetes Reduces Downtime</a> was originally published in <a href="https://medium.com/life-at-telkomsel">Life at Telkomsel</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>