Debunking the myth of serverless computing

Team had auto-scaling configured. Thresholds reviewed. Load balancer healthy. Full walkthrough done before the campaign. Traffic hit. 4 instances became 14. Response times went from 180ms to 4 seconds. Nobody questioned whether the application was actually the bottleneck. It was the most visible layer, most instrumented. Of course that is where you look first. The application was fine. It was just waiting. 14 connection pools hammering the same Azure SQL database. More compute, more pressure on the thing already struggling. This is what wrong assumptions cost in production. Not the incident. The 40 minutes spent looking in the right place with the wrong map. And this pattern shows up more than people think. The scaling reflex kicks in, the cache silently stops helping, the external rate limit does not care how many instances you are running. All of it traces back to the same thing: diagnosing symptoms instead of causes. This is part of The True Code of Production Systems, a series on decisions that only become visible when something breaks in production. Latest piece is a detailed breakdown of exactly how this plays out and what to look for instead. Link in the comments. #systemdesign #engineeringlessons #thetruecode

NEW BLOG POST! Two policies, three storage states, one rule connecting them. A newcomer's guide to caching, retention, and tiered storage in #MicrosoftFabric Eventhouse. https://lnkd.in/g5-Jc8Gs

🛑 Stop Polling, Start Reacting! ⚡ Before Azure Event Grid, if you wanted to process files the second they arrived, you had to run code 24/7. Your program would "knock on the door" of your storage every few seconds: "Is there a file? No. How about now? No." It was expensive, ineffective, and a massive waste of compute power. 💸 Enter the Storage Event Trigger powered by Event Grid: ✅ Zero Waste: Your pipeline stays 100% asleep until a file drops. ✅ Instant Action: The storage account "pushes" a signal the millisecond a file lands. ✅ Efficient Scaling: No more 24/7 idle servers. You only pay for the seconds your data is actually moving. It’s the difference between staring at your mailbox all day versus getting a notification on your phone. 📱 The Verdict: If you're still "polling" for data in 2026, it's time to switch to event-driven architecture. 🚀 #Azure #EventGrid #DataEngineering #CloudComputing #Serverless #TechEfficiency #AzureDataFactory https://lnkd.in/dEPFBpQr

## When a Unity Catalog (UC) Metastore is created, a specific external storage path is configured—typically pointing to an ADLS (Azure Data Lake Storage) location. This path acts as the root or boundary for data access within the metastore. All managed and external tables registered under the metastore must reside within this defined storage location or within explicitly authorized external locations. By default, access is strictly limited to these configured paths, meaning users or workloads cannot read from or write to any files or directories outside of them. To access additional data beyond the initial path, separate external locations must be defined and granted appropriate permissions. This ensures controlled, secure, and governed access to data across the environment.

What inspires me most about watching MCPize grow is the people behind it 🤗 . No one needs to be pushed 😅 . People come for their own reasons: to test their skills, learn something new, share their ideas, build something useful, or earn from what they create. And honestly, all of that deserves respect 👏 . Seeing new MCP servers appear on the platform, watching them evolve, and at the same time, seeing the platform itself grow and improve every day is exciting, motivating, and deeply inspiring 💫. And thank you to everyone curious, involved, supportive, and building with us. Step by step, this is becoming a real community of engaged people 💜 . https://mcpize.com/

0 → 500+ MCP servers in the marketplace. The number that surprised me wasn't the total. It was how fast the last 200 came in. A year ago MCPize was a landing page and a CLI. Now 500+ MCP servers are live: databases, browser automation, AI agents, web scrapers, payment tools. What changed: 85% revenue share. Developers build things when they get paid for it. Simple. The servers coming in now are more sophisticated. Less "hello world", more production-grade integrations that teams actually pay for. If you're building something with MCP, publish it. 500 in and there's still a lot of whitespace. #MCP #AI #buildinpublic

What good is AI-powered search if it goes down during a critical business moment? ❄   Snowflake just hit a major enterprise AI milestone: Cortex Search Service replication is now generally available.   You can now replicate your Cortex Search Services across accounts and regions — with full point-in-time consistency and native failover group support built right in.   No rebuilding indexes. No downtime. Your RAG chatbots, internal knowledge bases, and customer-facing search portals stay live, globally. 🔍   Think about what this means at scale: enterprises running AI-powered search across multiple regions can now handle outages or account migrations without breaking their user experience.   Is your enterprise AI search infrastructure truly resilient? With Snowflake Cortex Search replication, now it can be. 💡 For more details, See - https://lnkd.in/gUWUFhvz   #Snowflake #CortexAI #EnterpriseAI #DataCloud #AISearch

vLLM vs Triton Server An interesting discussion pinged through on my Reddit yesterday comparing vLLM and Triton Inference Server and trying to work out which is 'better'. The OP's thoughts were that vLLM is highly optimized specifically for LLMs, with features like KV caching and continuous batching making it particularly strong for transformer-based workloads. Triton, on the other hand, feels more like a general-purpose inference server which is more flexible, supporting multiple frameworks and model types. They also acknowledge that it's not really an apples-to-apples comparison as different use cases and purposes will likely make one 'better' than the other for that task. There has only been a few comments so far and they agreed that the use case is the biggest thing to which is more powerful and that generally speaking, vLLM is more dev-friendly for LLM serving, whereas Triton is more for the enterprise. Thought I would share the thread (link in comments) in case anyone wanted to jump in and get involved in the discussion.

Good Code Runs. Great Architecture Scales. If your system can’t handle real traffic, it’s not production-ready. Here’s what a scalable and secure architecture actually looks like: - VPC + Subnets (Foundation) - Public Subnet → Load Balancer (only entry point) - Private Subnet → App + Redis + DB Everything critical stays hidden inside a private network. - Load Balancer (Traffic Control) - Distributes traffic across instances - Eliminates single point of failure - Stateless App Instances (Scaling Power) - No session stored locally - Add/remove servers anytime - Redis (Session Management) - Centralized sessions across all instances - Seamless user experience - RDS (Secure Data Layer) - Runs in private subnet - No direct public access - S3 (Private Storage via VPC Endpoint) - Images/files stored securely - Accessed privately (no internet exposure) What you get: - Secure by design (private-first architecture) - Scales effortlessly (horizontal scaling) - Highly available (no downtime) - Reliable under real-world traffic Bottom line: Don’t just build features—build systems that survive scale. #AWS #SystemDesign #Scalability #CloudArchitecture #BackendEngineering

🚀 Top 10 Ways to Scale Systems Scaling isn’t just about adding servers—it’s about making the right design choices. Here are 10 proven ways to scale systems effectively 👇 ⚙️ 1. Vertical Scaling (Scale Up) • Increase CPU, RAM of a single machine ✔️ Simple ❌ Limited ceiling 🌍 2. Horizontal Scaling (Scale Out) • Add more servers ✔️ Highly scalable ✔️ Fault tolerant ⚡ 3. Load Balancing • Distribute traffic across servers ✔️ Prevents overload ✔️ Improves availability 💾 4. Caching • Store frequently accessed data (Redis, CDN) ✔️ Reduces DB load ✔️ Improves latency 🔄 5. Database Replication • Master → replicas ✔️ Scale reads ✔️ Improve availability 🧩 6. Database Sharding • Split data across multiple databases ✔️ Handles massive datasets ✔️ Improves write scalability 📦 7. Asynchronous Processing • Use queues (Kafka, RabbitMQ) ✔️ Decouples services ✔️ Handles spikes smoothly ⚡ 8. Rate Limiting • Control incoming traffic ✔️ Prevents system overload ✔️ Protects resources 🚀 9. Auto Scaling • Automatically add/remove servers based on load ✔️ Cost efficient ✔️ Handles traffic spikes dynamically 🌐 10. CDN (Content Delivery Network) • Serve content closer to users ✔️ Reduces latency ✔️ Offloads backend Which scaling technique has helped you the most? #SystemDesign #Scalability #DistributedSystems #BackendEngineering #Performance #Microservices