Failing Fast Saved Us From Failing Big
One of our internal services became unstable under load. We traced it to a slow downstream dependency with no timeout. The solution: Add circuit breakers with sliding window metrics Monitor P...
One of our internal services became unstable under load. We traced it to a slow downstream dependency with no timeout. The solution: Add circuit breakers with sliding window metrics Monitor P...
As Docker adoption grew, I started seeing teams treat images like ZIP files: messy, bloated, opaque. Key lesson: Your image is part of your contract. We built guidelines for: Slim base images ...
ThoughtWorks Radar (July 2014) listed Immutable.js — and it was a turning point for how I thought about frontend state. No more mutated state bugs. No more complex diffing logic. Immutable structu...
We started pushing more data through event logs this year. JSON was easy. Until it wasn’t. Problems: Extra fields broke consumers Enums misaligned across teams Schema changes became blocker...
Not every problem needs to be solved at scale. In 2014, we had a service that was underperforming. Everyone reached for: sharding, async queues, auto-scaling groups. But the actual fix? Rewrite a...
React was on the ThoughtWorks Radar this quarter — and for good reason. It felt like a breath of fresh air. No more tangled jQuery logic. No more DOM headaches. React’s mental model: UI as a p...
Java 8’s Streams API made pipelines readable. But as we explored async workloads, RxJava gave us even more control. Differences I felt: Streams = pull-based, synchronous Rx = push-based, asyn...
You wouldn’t deploy code without tests. So why deploy services without health checks? 2014 was the year we got serious about: Liveness checks (is it running?) Readiness checks (can it handle ...
ThoughtWorks Radar called out Logstash this quarter, and it deserved it. We adopted Logstash to clean and route logs from multiple systems: App logs Job queues Ingress/egress metrics The ...
I’ve seen microservice efforts collapse because teams couldn’t let go of shared databases. Here’s what went wrong: Coupled data models → fragile services Shared tables → ownership ambiguity ...