<?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 Samuel Joseph on Medium]]></title>
        <description><![CDATA[Stories by Samuel Joseph on Medium]]></description>
        <link>https://medium.com/@samuel-joseph?source=rss-1245436b7cf9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*Kf8Wet7tCkEPQNzm</url>
            <title>Stories by Samuel Joseph on Medium</title>
            <link>https://medium.com/@samuel-joseph?source=rss-1245436b7cf9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 05:20:26 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@samuel-joseph/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[How We Designed a Scalable AWS Architecture for High Availability]]></title>
            <link>https://medium.com/@samuel-joseph/how-we-designed-a-scalable-aws-architecture-for-high-availability-b592a1bb8890?source=rss-1245436b7cf9------2</link>
            <guid isPermaLink="false">https://medium.com/p/b592a1bb8890</guid>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[load-balancing]]></category>
            <category><![CDATA[autoscaling]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Samuel Joseph]]></dc:creator>
            <pubDate>Sat, 07 Mar 2026 09:04:04 GMT</pubDate>
            <atom:updated>2026-03-07T09:04:04.347Z</atom:updated>
            <content:encoded><![CDATA[<h3>𝐇𝐨𝐰 𝐖𝐞 𝐃𝐞𝐬𝐢𝐠𝐧𝐞𝐝 𝐚 𝐒𝐜𝐚𝐥𝐚𝐛𝐥𝐞 𝐀𝐖𝐒 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 𝐟𝐨𝐫 𝐇𝐢𝐠𝐡 𝐀𝐯𝐚𝐢𝐥𝐚𝐛𝐢𝐥𝐢𝐭𝐲</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rwUZUX5J1AMeU5EIFpJNEg.png" /></figure><h3>Most systems don’t fail because of traffic.</h3><p>They fail because the architecture was never designed for it.</p><p>I once reviewed an application deployed on 𝑨𝒎𝒂𝒛𝒐𝒏 𝑾𝒆𝒃 𝑺𝒆𝒓𝒗𝒊𝐜𝐞𝐬 that worked perfectly… until user growth doubled. Then the problems appeared:</p><p>• Intermittent downtime</p><p>• Unpredictable latency</p><p>• Slow scaling during peak traffic</p><p>The cloud wasn’t the problem. 𝗧𝗵𝗲 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝘄𝗮𝘀.</p><p>𝗧𝗵𝗲 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲: 𝑫𝒆𝒔𝒊𝒈𝒏 𝒇𝒐𝒓 𝑺𝒄𝒂𝒍𝒆 𝑬𝒂𝒓𝒍𝒚</p><p>We redesigned the infrastructure around one core assumption:</p><p>𝗧𝗿𝗮𝗳𝗳𝗶𝗰 𝘀𝗽𝗶𝗸𝗲𝘀 𝗮𝗿𝗲 𝗶𝗻𝗲𝘃𝗶𝘁𝗮𝗯𝗹𝗲. 𝗵𝗲𝗻𝗰𝗲, 𝗿𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝗰𝗲 𝗺𝘂𝘀𝘁 𝗰𝗼𝗺𝗲 𝗳𝗶𝗿𝘀𝘁.</p><p>The architecture needed to:</p><p>• Distribute traffic efficiently</p><p>• Scale automatically during demand spikes</p><p>• Remove single points of failure</p><p>• Maintain consistent latency</p><p>𝗟𝗼𝗮𝗱 𝗕𝗮𝗹𝗮𝗻𝗰𝗲𝗿 𝗧𝗿𝗮𝗱𝗲𝗼𝗳𝗳: 𝗔𝗟𝗕 𝘃𝘀 𝗡𝗟𝗕</p><p>One key decision was choosing between:</p><p>𝑨𝒑𝒑𝒍𝒊𝒄𝒂𝒕𝒊𝒐𝒏 𝑳𝒐𝒂𝒅 𝑩𝒂𝒍𝒂𝒏𝒄𝒆𝒓 (𝑨𝑳𝑩)</p><p>Best for application-aware routing.</p><p>Advantages</p><p>• Layer 7 HTTP/HTTPS routing</p><p>• Path-based routing for microservices</p><p>• WebSocket support</p><p>• Integration with authentication and WAF</p><p>𝙏𝙧𝙖𝙙𝙚𝙤𝙛𝙛: Slightly higher latency.</p><p>𝑵𝒆𝒕𝒘𝒐𝒓𝒌 𝑳𝒐𝒂𝒅 𝑩𝒂𝒍𝒂𝒏𝒄𝒆𝒓 (𝑵𝑳𝑩)</p><p>Built for high-performance networking.</p><p>Advantages</p><p>• Extremely low latency</p><p>• Handles millions of requests per second</p><p>• Static IP support</p><p>• Ideal for TCP/UDP workloads</p><p>𝙏𝙧𝙖𝙙𝙚𝙤𝙛𝙛: No application-level routing.</p><p>𝐎𝐮𝐫 𝐃𝐞𝐜𝐢𝐬𝐢𝐨𝐧</p><p>Because the platform was 𝙢𝙞𝙘𝙧𝙤𝙨𝙚𝙧𝙫𝙞𝙘𝙚-𝙗𝙖𝙨𝙚𝙙, we chose 𝘼𝙇𝘽.</p><p>We needed:</p><p>• Path-based routing (/api, /auth, /payments)</p><p>• HTTP-level observability</p><p>• Clear service segmentation</p><p>The routing flexibility outweighed the minor latency tradeoff.</p><p>𝐀𝐮𝐭𝐨𝐬𝐜𝐚𝐥𝐢𝐧𝐠 𝐒𝐭𝐫𝐚𝐭𝐞𝐠𝐲</p><p>A common issue in many deployments: Autoscaling triggers too late.</p><p>By the time new instances launch, the system is already struggling.</p><p>Instead, we implemented 𝗺𝘂𝗹𝘁𝗶-𝗺𝗲𝘁𝗿𝗶𝗰 𝗮𝘂𝘁𝗼𝘀𝗰𝗮𝗹𝗶𝗻𝗴 based on:</p><p>• CPU utilization</p><p>• Request count per target</p><p>• Application latency thresholds</p><p>This ensured scaling began before performance degraded.</p><p>𝐓𝐡𝐞 𝐅𝐢𝐧𝐚𝐥 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞</p><p>• Multi-AZ deployment</p><p>• Autoscaling groups across availability zones</p><p>• Stateless application containers</p><p>• Health checks replacing failed instances</p><p>𝐓𝐡𝐞 𝐫𝐞𝐬𝐮𝐥𝐭:</p><p>The system could absorb sudden traffic spikes without downtime.</p><p>Cloud platforms don’t magically guarantee scalability. Architecture does. That’s the difference between: 𝑫𝒆𝒑𝒍𝒐𝒚𝒊𝒏𝒈 𝒊𝒏 𝒕𝒉𝒆 𝒄𝒍𝒐𝒖𝒅 𝒗𝒔 𝑬𝒏𝒈𝒊𝒏𝒆𝒆𝒓𝒊𝒏𝒈 𝒇𝒐𝒓 𝒔𝒄𝒂𝒍𝒆.</p><p>If you’re building on AWS, ask yourself: 𝑫𝒊𝒅 𝒚𝒐𝒖 𝒅𝒆𝒔𝒊𝒈𝒏 𝒇𝒐𝒓 𝒕𝒐𝒅𝒂𝒚’𝒔 𝒕𝒓𝒂𝒇𝒇𝒊𝒄… 𝒐𝒓 𝒕𝒉𝒆 𝒕𝒓𝒂𝒇𝒇𝒊𝒄 𝒚𝒐𝒖 𝒉𝒐𝒑𝒆 𝒕𝒐 𝒓𝒆𝒂𝒄𝒉?</p><p>Curious to hear from other engineers: When would you choose NLB over ALB in production?</p><p>#AWS #CloudArchitecture #HighAvailability #DevOps #Autoscaling #InfrastructureDesign #CloudEngineering</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b592a1bb8890" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Top 5 Terraform Practices from Real World Projects]]></title>
            <link>https://medium.com/@samuel-joseph/my-top-5-terraform-practices-from-real-world-projects-d5623005d7e3?source=rss-1245436b7cf9------2</link>
            <guid isPermaLink="false">https://medium.com/p/d5623005d7e3</guid>
            <category><![CDATA[terraform]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Samuel Joseph]]></dc:creator>
            <pubDate>Sat, 24 May 2025 01:04:28 GMT</pubDate>
            <atom:updated>2025-05-24T01:04:28.990Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jUmteGyjqt6LCPyGkYHrLQ.png" /><figcaption>5 Terraform Hacks — Samuel Joseph</figcaption></figure><p>As a Senior Cloud and DevOps Engineer, I’ve used Terraform extensively across AWS, Azure, GCP, and even Oracle Cloud. Over the years, I’ve worked on greenfield builds, legacy migrations, and large-scale multi-cloud architectures. While every project is different, a few Terraform practices have stood out as critical for success; ensuring stability, scalability, and team collaboration.</p><p>Here are my top 5 Terraform best practices, learned (sometimes the hard way) from real-world projects.</p><h3>1. Use Remote State with State Locking</h3><blockquote>“Local state is fine — until it isn’t.”</blockquote><p>Storing Terraform state files locally is a recipe for drift and disaster when working in teams or automating with CI/CD. Always use remote state with locking mechanisms enabled.</p><p><strong>Preferred setup:</strong></p><ul><li>Use <strong>Terraform Cloud</strong>, <strong>AWS S3 + DynamoDB</strong>, or <strong>Azure Storage + Blob Locking</strong>. Locking prevents race conditions in concurrent terraform apply executions. It also provides visibility and collaboration for teams.</li></ul><p><strong>Example (AWS):</strong></p><pre>terraform {<br>  backend &quot;s3&quot; {<br>    bucket         = &quot;my-tf-state&quot;<br>    key            = &quot;dev/network/terraform.tfstate&quot;<br>    region         = &quot;us-west-2&quot;<br>    dynamodb_table = &quot;my-tf-locks&quot;<br>    encrypt        = true<br>  }<br>}</pre><h3>2. Modularize for Reusability and Consistency</h3><blockquote>“If you’re copying and pasting Terraform code, you’re doing it wrong.”</blockquote><p>Modules help reduce code duplication and make infrastructure reusable across environments or projects. I’ve created core modules for VPCs, security groups, IAM roles, and ECS clusters, then customized them via inputs.</p><p><strong>Benefits:</strong></p><ul><li>Faster provisioning.</li><li>Enforces organizational standards.</li><li>Easier onboarding for new engineers.</li></ul><p>Structure tip:</p><pre>modules/<br>  vpc/<br>  ecs/<br>  iam/<br>environments/<br>  dev/<br>  prod/</pre><p><strong>Keep your modules clean and version-controlled.</strong> Avoid over-engineering; start simple and refactor as you scale.3. Use Workspaces or Explicit Folder-Based Separation</p><p>For smaller projects, <strong>Terraform workspaces</strong> can help manage multiple environments (e.g., dev, qa, prod) using the same code base. However, for complex setups, I prefer <strong>explicit directory separation</strong>.</p><p><strong>Why folder-based separation often wins:</strong></p><ul><li>Easier CI/CD integration.</li><li>Less risk of accidental environment switches.</li><li>Better GitOps alignment.</li></ul><p>Example structure:</p><pre>live/<br>  dev/<br>    main.tf<br>  prod/<br>    main.tf</pre><p>Combine this with strong naming conventions and tagging strategies to avoid resource confusion.</p><h3>4. Enforce Code Quality and Policy-as-Code</h3><p>Terraform lets you define infrastructure as code, so treat it like software.</p><p><strong>Use tools like:</strong></p><ul><li>tflint and terraform fmt for style and syntax.</li><li>checkov or tfsec for security and compliance scanning.</li><li>Sentinel or OPA (Open Policy Agent) for policy-as-code.</li></ul><p>Integrate these checks into your CI/CD pipelines to catch issues early.</p><p>Example GitHub Actions snippet:</p><pre>- name: Terraform Format<br>  run: terraform fmt -check -recursive</pre><pre>- name: Run TFLint<br>  run: tflint --recursive</pre><h3>5. Tag Everything and Document Inputs/Outputs</h3><p>When you return to a Terraform project after 6 months, or when someone else does,<strong>tags and documentation</strong> save the day.</p><p><strong>Best practices:</strong></p><ul><li>Use consistent tagging across all resources (owner, environment, project, cost center).</li><li>Document module inputs and outputs with descriptions.</li><li>Use terraform-docs to generate README files for module</li></ul><p><strong>Example:</strong></p><pre>variable &quot;environment&quot; {<br>  description = &quot;The environment to deploy (e.g. dev, prod)&quot;<br>  type        = string<br>}</pre><pre>output &quot;vpc_id&quot; {<br>  description = &quot;The ID of the created VPC&quot;<br>  value       = aws_vpc.main.id<br>}</pre><h3>Final Thoughts</h3><p>These five practices; remote state, modularization, environment separation, code quality enforcement, and documentation, aren’t just <strong>“nice to have.”</strong> They’re foundational. They’ve helped me avoid downtime, scale teams, and build resilient cloud platforms.</p><p>If you’re starting out with Terraform, start with one or two of these and iterate. If you’re already deep into IaC, consider how your current workflows align with these principles.</p><p>What are your top Terraform lessons from the field? I’d love to hear them.</p><p>#Terraform #IaC #DevOps #CloudEngineering #InfrastructureAsCode #AWS #Azure #GCP #DevOpsBestPractices #CloudAutomation #TerraformModules #CICD #CloudInfrastructure #TerraformTips #DevOpsLife #TechLeadership</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d5623005d7e3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Your Cloud Strategy Keeps Failing (And How I Fixed It)]]></title>
            <link>https://medium.com/@samuel-joseph/why-your-cloud-strategy-keeps-failing-and-how-i-fixed-it-17baf3126379?source=rss-1245436b7cf9------2</link>
            <guid isPermaLink="false">https://medium.com/p/17baf3126379</guid>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[cloud-strategy]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[infrastructure-as-code]]></category>
            <dc:creator><![CDATA[Samuel Joseph]]></dc:creator>
            <pubDate>Mon, 12 May 2025 12:46:05 GMT</pubDate>
            <atom:updated>2025-05-12T14:03:02.433Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="failed cloud strategy" src="https://cdn-images-1.medium.com/max/1024/1*srpDRSfpIQgEJtCozKWaoQ.png" /><figcaption>Samuel Joseph — …failed cloud strategy…</figcaption></figure><p>Let’s be honest: most cloud “strategies” I come across aren’t really strategies; they’re glorified lift-and-shifts.</p><p>On paper, migrating to the cloud promises flexibility, scalability, and cost savings. But in reality? Many organizations just copy their on-prem setup into AWS, Azure, or GCP, and call it done.</p><p>As someone who’s spent years helping companies build scalable, secure, and cloud-native infrastructure across multi-cloud environments, I’ve seen this mistake over and over. But one project in particular made this painfully clear.</p><p><strong>The Lift-and-Shift That Didn’t Shift Anything</strong></p><p>I had just joined a mid-sized financial services company as a Cloud Transformation Lead. Leadership proudly told me they were “70% in the cloud.” That should’ve been a good sign, right?</p><p>But what I found was this:</p><ol><li>All their web, app, and database servers were provisioned on EC2 instances — <strong>configured almost exactly like their legacy setup.</strong></li><li>They managed access using manually configured Security Groups and Network ACLs —<strong> trying to mimic their old firewall rules.</strong></li><li>Scaling decisions were triggered manually: <strong>CloudWatch alarms were set to ping when CPU utilization went above 80% or dropped below 30%, at which point someone had to manually launch or terminate instances.</strong></li><li>All developers operated under IAM user credentials in a single shared AWS account — <strong>no role-based access, no federation, no real audit trail.</strong></li></ol><p>To them, this was “cloud.” To me, this was a rented data center — with all the old problems, now in a shinier interface. It wasn’t a cloud strategy. It was a technical translation with none of the cloud-native benefits.</p><h4><strong>How We Fixed It: From <em>LIFT and SHIFT </em>to <em>CLOUD NATIVE</em></strong></h4><p>I knew we had to shift our thinking — this wasn’t about fixing servers, it was about <em>fixing mindset, processes, and architecture. </em>Here’s how we turned it around:</p><h4>🔵 From Manual Scaling → Auto-Scaling</h4><ul><li>Replaced reactive CPU alerts with <strong>Auto Scaling Groups + intelligent policies</strong>.</li><li>Added <strong>Application Load Balancers</strong> for dynamic traffic distribution.</li><li>Result: Infrastructure scaled <strong>automatically</strong> — no human intervention needed.</li></ul><h4>🔵 From IAM Users → Federated, Role-Based Access</h4><ul><li>Integrated <strong>AWS SSO</strong> with their identity provider.</li><li>Moved to <strong>least-privilege IAM roles</strong> (mapped to teams/functions).</li><li>Result: <strong>Tighter security</strong>, easier audits, no more credential sprawl.</li></ul><h4>🔵 From ClickOps → Infrastructure-as-Code</h4><ul><li>Rewrote <strong>every security group, subnet, and alarm in Terraform</strong>.</li><li>Infrastructure became <strong>version-controlled, peer-reviewed, and auditable</strong>.</li><li>Result: No more “who changed what?” panic.</li></ul><h4>🔵 From EC2-Centric → Cloud-Native</h4><ul><li>Containerized apps → <strong>ECS + Fargate</strong> (no more patching EC2).</li><li>Let AWS handle scaling, provisioning, and maintenance.</li><li>Result: <strong>Faster deployments, lower ops overhead</strong>.</li></ul><h3><strong>The Results</strong></h3><p>Deployment times dropped from weeks to under an hour</p><p>◼️Scaling became predictable and automatic</p><p>◼️Developers no longer fought over IAM policies or waited for infra</p><p>◼️We cut compute costs without sacrificing performance</p><p><strong>Most importantly; </strong>The cloud stopped being someone else’s job. It became part of how every team delivered value.</p><h4><strong>Final Thought</strong></h4><p>Your cloud strategy isn’t failing because the cloud is broken. It’s failing because you’re treating the cloud like your old data center.</p><p>If you’re still managing users manually, reacting to CPU alarms, and spinning up EC2s for every new service — you haven’t transformed. You’ve just migrated your problems.</p><p>I’ve helped lead this change. I’ve lived through the mess. And I can tell you this: <strong><em>“Real cloud transformation happens when you stop lifting and start rethinking.”</em></strong></p><p><strong>Let’s Connect</strong><br> I’m Samuel Joseph, and I help companies turn their cloud chaos into clarity. If you’re stuck in a costly, rigid, or outdated cloud setup, let’s talk.</p><p>#DevOps #CloudComputing #CloudStrategy #AWS #Azure #GCP #Terraform #Leadership</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=17baf3126379" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>