Tech Migration Best Practices

Over the weekend, I read Google's paper on how they use AI for internal code migrations—and it’s packed with insights on how to approach legacy system modernization. I’ve attached the paper for those interested, but here’s how I believe some of these strategies can help us tackle complex modernization challenges: 🔎 1. Accelerating Legacy System Modernization Google leverages Large Language Models (LLMs) to automate large-scale code migrations, significantly reducing manual effort and speeding up projects. Applying similar AI-driven approaches can streamline the modernization of legacy systems, cutting through complexity and outdated code. 🔎 2. Combining AI with Proven Engineering Tools By blending LLMs with Abstract Syntax Tree (AST)-based tools, the ensure accuracy and scalability in their code transformations. This hybrid method shows how AI and traditional engineering techniques can work together to deliver safe and reliable modernization. 🔎 3. Reusable Migration Workflows Google created modular, reusable workflows that make onboarding and executing new migration tasks faster and more efficient. Developing similar toolkits for legacy systems could simplify recurring modernization steps and adapt to complex scenarios. 🔎 4. Measuring Success by Business Impact Google focuses on measurable outcomes, like a 50% reduction in project time, rather than just the volume of AI-generated code. This business-aligned metric highlights the importance of demonstrating clear ROI in technology transformation projects. 🔎 5. Safe and Scalable Rollouts Their phased deployment strategy ensures AI-driven changes are rolled out safely, minimizing disruption. Adopting a controlled rollout approach can help manage risks and ensure stability when modernizing critical systems. 🔎 6. Strategic Use of AI Models Google balances using custom fine-tuned models and general-purpose tools depending on the task. This approach offers valuable insight into when to invest in specialized AI solutions versus using adaptable off-the-shelf models. 📌 The Big Picture: Legacy system modernization is about combining AI-driven efficiency with engineering best practices to deliver faster, safer, and more impactful business transformations. 📎 I’ve attached the paper if you’d like to explore it further! #LegacyModernization #GenAI #BusinessInnovation — Enjoyed this post? Like 👍, comment 💭, or repost ♻️ to share with others.

I constantly hear shocking stories of cloud migration mistakes that spiral into unexpected, skyrocketing costs beyond what anyone ever imagined. Most companies underestimate the complexity. Skip dependency mapping. Pay the price. Cloud migrations go beyond moving workloads - they require knowing what to move, when, and how it affects the rest of your environment. Without a solid plan, you risk unplanned downtime, security gaps, and overspending on misconfigured cloud resources. Here’s how to migrate without chaos: 1. Start with full visibility. Map every application, service, and dependency before migration. Unknown connections lead to downtime, security risks, and hidden costs. Many organizations don’t realize how interconnected their systems are until something breaks. 2. Assess workloads before moving them. Not everything belongs in the cloud. Classify applications by criticality, complexity, and cloud readiness. Legacy systems often need refactoring or special configurations, while certain workloads may be better off staying on-premises. 3. Move in phases, not all at once. A "lift and shift" migration can break critical systems. Migrate in controlled stages, test thoroughly, and adjust before moving forward. Pilot test with non-critical workloads first, gather insights, then move mission-critical systems. 4. Optimize before the migration. Unused resources drain your budget. Right-size workloads, eliminate redundant services, and continuously monitor costs. Cloud sprawl - where forgotten instances keep running - can waste thousands per month. 5. Avoid compliance blind spots. Migrating nodes without visibility can lead to regulatory violations and security gaps. Ensure sensitive workloads follow security best practices before, during, and after migration. The hard truth? You can’t migrate what you don’t know about. Map -> Plan -> Migrate. NO SHORTCUTS.

NEW MIGRATION GUIDANCE - Cloud migrations can be complex, but they don’t have to be uncertain, whether you're moving from on-premises environments or other clouds. To help bring more clarity, we published new Cloud Migration guidance in Microsoft’s Cloud Adoption Framework. This guidance offers a structured roadmap for migrating workloads to Azure from both on-premises and other cloud platforms. It’s the result of close collaboration with Microsoft experts and Microsoft MVPs. It reflects lessons learned from thousands of real-world migrations. The goal is to support teams at any stage of their cloud journey with clear, actionable steps.   Migration Process Overview: 1️⃣ Plan Your Migration 1. Assess readiness and team skills 2. Choose data migration paths 3. Define migration sequencing and rollback plans 4. Engage stakeholders 2️⃣ Prepare Workloads for the Cloud 1. Fix compatibility issues 2. Validate workloads' functionality 3. Build reusable infrastructure 4. Document deployment steps 3️⃣ Execute Migration to the Cloud 1. Prepare stakeholders and freeze changes 2. Finalize production environment 3. Execute cutover and validate success 4. Provide stabilization support 4️⃣Optimize Workloads After Migration 1. Fine-tune configurations in the cloud 2. Collect and act on user feedback 3. Review workloads regularly 4. Optimize hybrid and multicloud dependencies 5️⃣Decommission Source Workloads 1. Confirm decommissioning with stakeholders 2. Reclaim or reassign licenses 3. Preserve data for compliance 4. Update documentation and architecture records 🔗 Explore the new migration guidance here: https://lnkd.in/e2VgCU8m If you're navigating a cloud migration or supporting those who are, I hope this provides the guidance you need. 📣 Acknowledgments: This work reflects the contributions of many across the Microsoft community:   Microsoft MVPs: Stéphane Eyskens, Michael Stephenson, Danny McDermott, Stanislav Zhelyazkov, Joe Carlyle, Scott Corio, Simon Wåhlin, Bert Wolters, Elton Bordim, Haiko Hertes, Robert Hogg, Vladimir Stefanović, Andrew Wilson   Microsoft colleagues: Daniel Söderholm, Ivan Bondy, Rob Rinear, Brody Schulke, Philip Sills, Sandra Patricia Sánchez Martínez, Jack Tracey, Sunil Seth, Timo Salomäki, Michael Lemire, Tomas Kovarik, Larz Stridh, Konstantinos Pantos, Ryan Pfalz, Oscar Zamora, Courtney Taylor, PMP, Kevin Bell, John Lunn, Mannan Mohammed, Mark Piggott, Phani Kumar Teluguti, Yudhbir Singh, Alvaro Guadamillas Herranz CAF Engineering Lead: Jason Bouska Luke Nyswonger, Martin Ekuan, Hans Yang

We migrated 60+ microservices to Kubernetes in 4 weeks. And we threw half the “best practices” out the window. Here’s what actually worked (and what didn’t): 1. 𝐃𝐨𝐧’𝐭 𝐨𝐯𝐞𝐫-𝐚𝐛𝐬𝐭𝐫𝐚𝐜𝐭. We ditched Helm for some services. Why? The templates were bloated. Debugging was a nightmare. For core infra services, raw YAML + Kustomize gave us more control, less magic. 2. 𝐒𝐤𝐢𝐩 𝐭𝐡𝐞 𝐂𝐈/𝐂𝐃 𝐩𝐞𝐫𝐟𝐞𝐜𝐭𝐢𝐨𝐧 𝐭𝐫𝐚𝐩. We didn’t build fancy pipelines upfront. We used kubectl apply with Git-backed manifests for the first two weeks. Speed > Purity. Once stable, we baked in GitOps flows via ArgoCD. 3. 𝐅𝐨𝐫𝐠𝐞𝐭 𝐟𝐮𝐥𝐥-𝐬𝐜𝐚𝐥𝐞 𝐦𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠 (𝐢𝐧𝐢𝐭𝐢𝐚𝐥𝐥𝐲.) We didn’t set up all dashboards on Day 1. Just kubectl top, resource limits, and simple health probes. Only after stabilizing did we plug in Prometheus, Grafana, Loki. 4. 𝐍𝐚𝐦𝐞𝐬𝐩𝐚𝐜𝐞𝐬 > 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐬. Everyone says “multi-cluster” is the way. We found namespaces + strong RBAC + network policies was faster, safer, and cheaper to ship fast. 5. 𝐃𝐞𝐟𝐚𝐮𝐥𝐭 𝐥𝐢𝐦𝐢𝐭𝐬 𝐰𝐢𝐥𝐥 𝐡𝐮𝐫𝐭 𝐲𝐨𝐮. We got bitten by K8s defaults. Some pods OOMed. Others ran wild. Set memory/cpu requests+limits for every service. Don’t assume your team will remember. 6. 𝐎𝐧𝐞 𝐭𝐞𝐚𝐦, 𝐨𝐧𝐞 𝐩𝐥𝐚𝐲𝐛𝐨𝐨𝐤. We wrote a 12-page internal doc on service onboarding. No Slack chaos. No tribal knowledge. Everyone knew how to get from local → dev → prod. 7. 𝐌𝐨𝐯𝐞 𝐟𝐚𝐬𝐭, 𝐛𝐮𝐭 𝐛𝐚𝐜𝐤 𝐢𝐭 𝐮𝐩. We snapshot’d etcd + backed up persistent volumes daily. We never needed them. But they let us sleep. 𝐌𝐨𝐬𝐭 𝐊8𝐬 𝐦𝐢𝐠𝐫𝐚𝐭𝐢𝐨𝐧𝐬 𝐟𝐚𝐢𝐥 𝐛𝐞𝐜𝐚𝐮𝐬𝐞 𝐭𝐞𝐚𝐦𝐬 𝐨𝐛𝐬𝐞𝐬𝐬 𝐨𝐯𝐞𝐫 𝐭𝐨𝐨𝐥𝐢𝐧𝐠 𝐚𝐧𝐝 𝐢𝐠𝐧𝐨𝐫𝐞 𝐭𝐞𝐚𝐦 𝐚𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭. We shipped in 4 weeks not because we followed the rules. But because we knew exactly which ones to break. If you’re planning a K8s migration, I’ll happily share our internal playbook. Just comment “K8s” and I’ll DM it.

The next few years are going to be tough. Many legacy applications finally need to be modernized.  10 actions to survive. 1. Focus: Not every functionality needs to be migrated. Strict scope management based on real customer needs is crucial. What's your approach to scope prioritization? 2. Outcome-driven: Delivered functionality isn't the main success criterion - improved business value is. In my last project, we delivered 18% more revenue with just 60% of the migrated functionality. What metrics matter most in your modernization efforts? 3. Data-driven: Validate the value of each delivered feature through A/B testing. Combine quantitative data with user stories to paint the complete picture. 4. Incremental and iterative: From month one, deploy continuously to production through a robust delivery pipeline. Daily releases should be your minimum target. Agile and DevOps work. 5. Fail fast: Build and validate technically risky and commercially important functionalities first. Minimize basic functionality. Effectiveness before efficiency. 6. Experience-based: Don't reinvent the wheel. Learn from others who've succeeded. Shamelessly adopt state-of-the-art practices that work. 7. Human-centric: Your employees are critical to success. They understand customer needs, business processes, and legacy systems. Blend their experience with external expertise and invest in change management. 8. Be adaptable: We plan, God laughs. Observe, reflect, and adapt regularly at every organizational level. Stay self-critical and embrace change. 9. Cost-aware: Modernization isn't just about technology - it's about business value. Track and communicate both investment and returns. Create transparency about technical debt reduction and new revenue opportunities. 10. Future-proof: Design for change, not just today's requirements. Choose modern, maintainable architectures and build technical excellence into your culture. Microservices aren't dead. Which of these measures resonates most with your experience? What would you add to this list? Share your thoughts in the comments!

I have never considered it a wise decision to switch technologies just because the other one looks “prettier”. Choices like that rarely end well. There’s a clear difference between being forced to rewrite a system and switching purely out of trend. If you have a prototype built in Java, Python or any other language just to validate an idea, and once that idea proves viable and needs to scale, migrating to something more robust makes perfect sense — I’ve been through this in critical MarketData scenarios, where I had to rewrite a Java prototype in C++ to meet strict performance requirements. In these cases, migration is absolutely the right call. I’ve also had to migrate applications running on macOS servers to Linux, reducing infrastructure costs by an astonishing 97% — a strategic move that made perfect business sense. These are the kinds of migrations that truly add value: they solve real problems, improve scalability, and save money, rather than creating new headaches. But abandoning a solid system, written in a language that is perfect for that type of domain, with an active community and constant updates, just to adopt another because it is “prettier” or “supposedly safer”, is a classic mistake. The outcome? Always the same! A flood of new bugs, customer churn, delayed projects, blown budgets, damaged reputation, and an unnecessary stress load for the entire team. Switching technologies just for fashion is trading productivity for chaos! Fabio Galuppo, Leidson Campos A. Ferreira, José Augusto N. G. Manzano Thread suggested by Fabio Galuppo #TechLeadership #SoftwareArchitecture #EngineeringDecisions #CodeQuality #Scalability #MarketData #CPlusPlus #SoftwareDevelopment #EngineeringWisdom #DontChaseTrends

From my recent S/4HANA Experience, a key topic discussed was SAP ECC to S/4HANA Migration – Choosing the Right Path. Moving from ECC to S/4HANA is more than an upgrade. It is a strategic transformation. The migration should be guided by business priorities first, with technology as the enabler. Efficiency, agility and cost control must shape the path before system configurations come into play. Greenfield (New Implementation): A complete rebuild on S/4HANA, migrating only essential master data while leaving behind old transactions and customizations. Best suited for organizations seeking simplification and modernization. Example: Re-designing material planning processes with fresh MRP logic and analytics, free from years of complexity. Brownfield (System Conversion) Direct conversion of ECC to S/4HANA, retaining history, custom developments and configurations. Ideal for continuity with minimal disruption. Example: Preserving long-term plant maintenance history while enabling faster analytics. Hybrid (Selective Data Transition) A mix of both, where some modules are re-implemented and others converted as is. Practical for phased modernization in large-scale environments. Example: Finance and procurement rebuilt with automation, while warehouse operations transition gradually.

Data migration is not just “move data from A to B.” It is a system design decision. The wrong migration pattern can create downtime, broken pipelines, data mismatches, reporting gaps, and rollback panic. Modern data engineering teams need to choose the pattern based on risk, scale, system complexity, and business tolerance for disruption. Here are the 7 data migration patterns every modern data engineering team should know: 𝗟𝗶𝗳𝘁-𝗮𝗻𝗱-𝗦𝗵𝗶𝗳𝘁 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Move existing data, jobs, and systems to a new environment with minimal redesign. Best when speed matters more than modernization. 𝗥𝗲-𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺𝗶𝗻𝗴 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Move to a new platform while making small improvements to performance, cost, tooling, or scalability. 𝗥𝗲-𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Redesign core data flows for better scalability, reliability, and long-term flexibility. 𝗣𝗵𝗮𝘀𝗲𝗱 𝗖𝘂𝘁𝗼𝘃𝗲𝗿 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Move users, workloads, datasets, or teams in controlled phases instead of switching everything at once. 𝗗𝘂𝗮𝗹-𝗪𝗿𝗶𝘁𝗲 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Write data to both old and new systems during the transition so outputs can be compared before retiring legacy systems. 𝗖𝗗𝗖 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Use Change Data Capture to replicate inserts, updates, and deletes from the source system to the target in near real time. 𝗥𝗼𝗹𝗹𝗯𝗮𝗰𝗸-𝗥𝗲𝗮𝗱𝘆 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻 Design the migration so the team can safely return to the old system if the new one fails. The best migration teams do not just plan the happy path. They plan validation, reconciliation, monitoring, failure detection, rollback, and stakeholder impact. Because in data migration, success is not only moving data. Success is moving trust. Which data migration pattern have you used most often? Follow Sumit Gupta 📊 for more such insights!!

One mistake I see in cloud migrations over and over is that teams lift-and-shift apps into cloud VMs and then wonder why costs and security get worse. Here's what actually matters. In many cases, people do just a lift-and-shift. They pick up an application, figure out a way to run it in a VM, move it over into the cloud, and forget it. But unless they do a business impact assessment and a technology architecture review, they’re probably going to end up with higher costs, worse security, and an application that may not meet current business needs. I’ve done it myself. When I was CIO at Microsoft, my goal was to get out of our data centers and move as much as I could to Microsoft Azure. Some apps we re-architected. Others, we just parked in VMs and said, “We’ll get to it next year.” That wasn’t optimal, but at the time, it let us vacate data centers and retire old equipment. The problem isn’t that lift-and-shift is generally wrong, but when leaders treat it as the default strategy, instead of a short-term fix with a plan behind it. Here’s how I’ve learned to approach it: 1. Inventory and prioritize your apps. Not everything is mission-critical, but not everything should be punted either. 2. Do a business-impact assessment and a technology architecture review. Without these, you risk higher costs and worse outcomes. 3. Use specialists, not generalists. Moving apps is like redoing plumbing. You can ask me to figure it out, but it’ll take me three times as long, and it won’t look good. And, if not properly performed, can lead to disaster. It’s worth it to have people who know the craft. 4. Allow temporary “parking” in VMs* but set a clear timeline to revisit, and then force a decision to terminate or optimize for cloud. Short-term convenience without discipline creates long-term debt. Costs rise, risks increase, and you may be burdened with apps that don’t do what the business needs.

We interviewed data leaders who've been through major platform migrations, and one thing became crystal clear: migrations are so critical to the data team’s ability to impact the business that they define the data team’s success and data engineer's careers. The migration I ran at Lyft certainly defined mine — and put me on a path to build Datafold. Migrations are a perfect storm of high stakes, tight deadlines, and zero room for error. You're not just moving data – you're juggling pipeline re-engineering, data validation, stakeholder management, and keeping production running. The insights from these leaders were surprisingly consistent: > Lift-and-shift first, optimize later. The temptation to fix everything during migration is strong but often creates more problems than it solves. > Stakeholder trust is everything. Without clear proof of data parity, skepticism creeps in, systems run in parallel longer than planned and costs balloon. > The "last mile" is where migrations go to die. When you think you're done, stakeholder reviews uncover edge cases that trigger new refinement cycles. But here's what gives me hope: we're finally at a technological inflection point. What used to take years can now be done in months., and whatrequired armies of engineers can now be automated. We've compiled these stories, lessons, and a practical framework for modern migrations in our latest guide: https://lnkd.in/em-BBqM4