Interested in how we test systems at Amazon?
In a new article in Communications of the ACM, Ankush and I write about the evolution of systems correctness and testing practices at AWS. We cover classic and formal approaches, and new ones like deterministic simulation.
cacm.acm.org/practice/sys...
Current London 2025: Kafka, Flink, Iceberg, AI... and some drama.
www.streamingdata.tech/p/current-lo...
also spotify
As reliable as Google Drive is, when viewed as an external system, it's a single point of failure. That's why we backup all data that's stored there - Jakub shares our approach, accompanied by an OSS project!
softwaremill.com/how-we-appro...
also, do you have know of any benchmarks/articles comparing of the performance of partitioned vs non-partitioned tables?
That's a really great feature from Postgres. I've so many questions: If you do a select from a table do you need to specify the partition or can you just reference the table? What happens if you want to query multiple partitions - does postgres handle this automatically?
Replace "github" with "gitingest" in the url, and you get the whole repo as a single string that you can then paste in your LLMs
Building Bluesky: a Distributed Social Network - in @pragmaticengineer.com's Real-World Engineering Challenges series
newsletter.pragmaticengineer.com/p/bluesky
What's the advantage of Guava's monitor/guard have over the standard java libraries primitives? Guava have some nice collections and functions over these collections, but when it comes to concurrency I usually stick with the java SDL primites.
Some images from the Twitter years in a thread (since they are no longer accessible) 1/n
