<?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[Stories by Kiran Mutukuru on Medium]]></title>
        <description><![CDATA[Stories by Kiran Mutukuru on Medium]]></description>
        <link>https://medium.com/@kiran.mutukuru?source=rss-3c901924002f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*F7BuVNrGjZaXsMn9F1D93g.jpeg</url>
            <title>Stories by Kiran Mutukuru on Medium</title>
            <link>https://medium.com/@kiran.mutukuru?source=rss-3c901924002f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 12:57:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kiran.mutukuru/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[AI: my Top Ten]]></title>
            <link>https://medium.com/@kiran.mutukuru/ai-my-top-ten-05b00cb2e324?source=rss-3c901924002f------2</link>
            <guid isPermaLink="false">https://medium.com/p/05b00cb2e324</guid>
            <dc:creator><![CDATA[Kiran Mutukuru]]></dc:creator>
            <pubDate>Sat, 20 Apr 2024 00:05:18 GMT</pubDate>
            <atom:updated>2024-04-25T11:58:24.295Z</atom:updated>
            <content:encoded><![CDATA[<p>If you are like me, you are tired of hearing about the rate at which user base grew for Chat GPT. If you are like me, you poured over various materials and applications for Generative AI and for AI in general in the past few months. Here are my top ten observations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6jZF1_GhqlHjze4S-TF1SA.png" /><figcaption>AI, ML &amp; friends</figcaption></figure><ol><li>Artificial Intelligence is the biggest disruptor of technology, business, and society in a generation. Technology leaders should embrace the application of AI into solutions. Some are no-brainers like employing chatbots for customer service. Businesses differentiate by using AI in their core business functions.</li><li>Machine Learning has been around. The advancements in computer vision, natural language processing, reinforcement learning, and the application of deep learning made the step change of Generative AI possible.</li><li>AI is an application on top of data. An LLM throws processing power on top of bulk data. It won’t get very far if your data is garbage.</li><li>Gen AI will push industries to cloud adoption. On the flip side, Small LMs will be required to generate AI on-premises.</li><li>AI gives a low barrier of entry for bad actors. Secure organisations fight bad AI with good AI.</li><li>Generative AI is more than search that retrieves essay answers. Multi-modal gen AI, and AI agents will transform this capability multi-fold with tremendous business potential. Personal use of AI is fun but building enterprise AI is better.</li><li>AI must be explainable and transparent to the users to make it ethical and valuable.</li><li>Sustainability of AI solutions should be at the forefront. The amount of energy spent by an AI provider for a gen AI transaction should be published (like nutritional facts of packaged food).</li><li>Society leaders must understand and regulate AI where they can. The amount of change to be managed in the coming years is great. The promise of prompt engineering as a profession is not sufficient to balance the broad work displacement.</li><li>The G in AGI, if achievable, is not possible for a long time to come. The confluence of AI and the ‘hard problem of consciousness’ are welcome areas to be explored.</li></ol><p>AI is here to stay irrespective of how the magnificent seven do in the stock market. Share your views in the comments section.</p><p><em>#GenAI #EthicalAI #AIonCloud #AIonPrem #DataForAI #Consciousness #watsonx</em></p><h4>References (subset)</h4><p>1. Martin Keen, “The most important AI trends in 2024”, IBM Technology, <a href="https://youtu.be/sGZ6AlAnULc?si=KSaHO1XnN9o4kyw6">https://youtu.be/sGZ6AlAnULc?si=KSaHO1XnN9o4kyw6</a></p><p>2. Kirill Eremenko, “Artificial Intelligence (AI) for Executives and Top-Level Managers: understand Artificial Intelligence in under 2 Hours”, Udemy 2024</p><p>3. Mike Wooldridge, “What is the future of Generative AI — The Turing Lectures”, The Royal Institution, <a href="https://youtu.be/b76gsOSkHB4?si=3TGC_PqmzZBRnThp">https://youtu.be/b76gsOSkHB4?si=3TGC_PqmzZBRnThp</a></p><p>4. Swami Sarvapriyananda, “Reality and Reality+”, <a href="https://youtu.be/WWca-rhjnro?si=vOFnJm-dB5i442f-">https://youtu.be/WWca-rhjnro?si=vOFnJm-dB5i442f-</a></p><p>5. Jon Stewart, “The false promises of AI”, <a href="https://youtu.be/20TAkcy3aBY?si=f9PcIv3TVCW4Zqdm">https://youtu.be/20TAkcy3aBY?si=f9PcIv3TVCW4Zqdm</a></p><p>6. Taimur Ijlal, “Generative AI: Risk and Cyber Security Masterclass”, Udemy 2024</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=05b00cb2e324" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Continuous Availability — a primer]]></title>
            <link>https://medium.com/@kiran.mutukuru/continuous-availability-a-primer-87b414d11ff4?source=rss-3c901924002f------2</link>
            <guid isPermaLink="false">https://medium.com/p/87b414d11ff4</guid>
            <dc:creator><![CDATA[Kiran Mutukuru]]></dc:creator>
            <pubDate>Wed, 10 May 2023 00:56:48 GMT</pubDate>
            <atom:updated>2023-05-10T00:56:48.012Z</atom:updated>
            <content:encoded><![CDATA[<h3>Continuous Availability — a primer</h3><p>A resilient application is continuously available to the users, irrespective of any failures in its technology stack. A user perceives availability in terms of how accessible his or her data is, how performant the app is, and how reliable his or her experience is. A software developer may consider the robustness of error handling as a key factor of application availability. An IT support person views Availability as how “up” the systems are. Markets are not kind to availability issues faced by customers.</p><p>To design a continuously available application, one must not only consider the disaster recovery of its infrastructure, but the end-to-end stack related to each customer experience path provided by the application. This article is meant to be a primer for defining the scope of availability of IT applications.</p><h3>Key Metrics</h3><p>The number of 9s: how close an application is to being 100% available.</p><blockquote>Two 9s or 99% available implies a downtime of almost 4 days in an year. Four 9s means 99.99% or a downtime of 52 minutes. Six 9s means 99.9999% or a downtime of less than a minute.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/588/1*a6sgbubmxVNa77rOIbVuRg.png" /><figcaption>Figure 1: More 9s, Less Downtime</figcaption></figure><p>The difference in impact between Two 9s and Six 9s is difference between users out of your application for a long weekend versus a minor glitch in their experience. Zero-downtime is the ideal which theoretically refers to infinite number of 9s in the architecture. Note that availability comes at an expense — infrastructure costs, observability &amp; SRE (Site Reliability Engineering) costs, among others. It is easy to imagine an exponential cost curve that is inverse of the outage curve shown above.</p><p>Enterprises plan Disaster Recovery (DR) to address how quickly and how gracefully they can bring back the applications after an unplanned outage. Once a disaster strikes, i.e., some kind of disruption to any component in the architecture, enterprises must take action to minimise the downtime to the users.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/644/1*oBJ2tnlMmLS2TIczoItf_g.png" /><figcaption>Figure 2: Recovery Objectives</figcaption></figure><blockquote>Recovery Time Objective (RTO): the time it takes for an application to be available after a disaster event. The shorter the RTO the better.</blockquote><blockquote>Recovery Period Objective (RPO): the point of time before a disaster event up to which the application can rollback to, in terms of the data retrieved. The shorter the RPO the better.</blockquote><h3>Availability Zones</h3><p>Availability Zones (AZ), or Data Centers are geographically separated to accommodate high availability, for DR and to cater to geographical proximity of your userbase. Applications are denoted as Active-Active or Active-Passive to denote if both zones can be accessed by users simultaneously for real-time data consumption.</p><p>An Active-Active application is reachable by consumers in both availability zones — and provides consistent data from both AZs. In other words, there is no difference in application behavior if dependent applications access it in one AZ or the other. A setup where the application is actively running in 3 AZs (as shown in the diagram below) is sometimes referred to as “3-Active”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/378/1*-H1nZESh0cCZQSeg5FjDLw.png" /><figcaption>Figure 3: Active in 3 sites</figcaption></figure><p>A 3-Active application allows for a DR setup ready to go for additional DR situations, and for robust release management. In industries where cloud adoption has been deliberately slow, organizations keep two AZs on-premises, and use a 3rd zone that employs cloud for analytics use cases.</p><h3>System Context of Continuous Availability</h3><p>As already stated, availability is not confined to infrastructure. Availability must be considered during design and throughout the software life cycle. An indication of various aspects is shown in the figure below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/660/1*kndMI-PgPdaRTEh1q3fKaA.png" /><figcaption>Figure 4: System Context for Availability</figcaption></figure><h4>Design and Development</h4><p>The CAP theorem states that a distributed application can deliver only two of the three desired characteristics: Consistency, Availability, and Partition Tolerance. Among the ability for applications to view similar set of data when reaching different nodes, the ability to get the data even when one or more nodes are down, and the ability for an application cluster to continue to work despite communication breakdowns, we can only achieve two abilities. An application that performs poorly affects the perception of availability. End-to-end security is paramount concerns for a continuously available enterprise application.</p><h4>Deployments</h4><p>CI/CD processes, people and tools must consider the Availability architecture to automate deployments appropriately between data centers (or AZs). There may be certain components not actively running in a secondary passive data center. Applications must “know” their local data center for low latency integrations with other applications.</p><h4>Chaos Testing</h4><p>In addition to diligent traditional testing practices of unit testing, integration testing, regression testing, enterprises are embracing the use of Chaos testing. It is the concept of throwing a wrench into perfectly functional applications, to proactively identify problems. Some of the concepts also fall under DR testing where individual components of tech stack are visualized to be down and observing application behavior with the goal of fixing any problems thus observed.</p><h4>Infrastructure-as-a-service</h4><p>Cloud practitioners love the acronym, IaaS. Here, I use it to abstract virtual machines and underlying hardware, whether on-premises or on cloud. Planning around network latency between data centers, and ensuring sufficient capacity in individual data centers, Monitoring replication technologies, and switching software components between data centers quickly are all chief concerns. Load balancing important business functions between data centers, typically to cater to specific geographies in steady state is also an important aspect in achieving continuous availability.</p><h4>Application Monitoring</h4><p>APM (Application Performance Monitoring) is critical to a continuously available application. From sophisticated reporting of performance metrics to synthetic monitors that notify support staff upon failures of components, and to the ability to collate tracing data between data centers are all key features.</p><h4>Disaster Recovery</h4><p>Disasters come in many sizes and shapes. DR planning requires discipline about understanding your tech stack and identifying potential pitfalls, even if the failure scenarios seem theoretical. It is also essential to identify the scope and quantity of data loss possible upon a disaster. Runbooks that list out the steps of corrective action to re-establish a cluster or a data center, with automation are critical to achieving small RTO and RPO.</p><p>If you are responsible for enterprise applications, what keeps you up at night? Let me know!</p><p><em>#resiliency #continuous-availability #AlwaysOn #HADR</em></p><h4>References</h4><p>1. Availability in a Cloud-native world, <a href="https://www.ibm.com/cloud/architecture/architecture/practices/achieve-zero-downtime">https://www.ibm.com/cloud/architecture/architecture/practices/achieve-zero-downtime</a>, Retrieved in February 2020</p><p>2. Shared Responsibility Model, <a href="https://docs.aws.amazon.com/whitepapers/latest/disaster-recovery-workloads-on-aws/shared-responsibility-model-for-resiliency.html">https://docs.aws.amazon.com/whitepapers/latest/disaster-recovery-workloads-on-aws/shared-responsibility-model-for-resiliency.html</a>, Retrieved in April 2023</p><p>3. Patterns for resilient apps, <a href="https://cloud.google.com/architecture/scalable-and-resilient-apps">https://cloud.google.com/architecture/scalable-and-resilient-apps</a>, Retrieved in May 2023</p><p>4. Chaos Testing, <a href="https://www.lambdatest.com/learning-hub/chaos-testing#chaos-testing-pyramid">https://www.lambdatest.com/learning-hub/chaos-testing#chaos-testing-pyramid</a>, Retrieved in May 2023</p><p>5. CAP theorem, <a href="https://www.ibm.com/topics/cap-theorem">https://www.ibm.com/topics/cap-theorem</a>, Retrieved in May 2023</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=87b414d11ff4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>