The Network Latency Tax (on your build cache)
summary
Bitrise builds a global, CDN-backed build cache with a Gradle plugin that uses gRPC streaming over HTTP/2, putting a primary cache in-datacenter and replicas at edge nodes, plus hot in-memory storage backed by SSD and content-defined chunking for dedupe. Real numbers: intercontinental RTT ~150 ms, in-datacenter ~0.5 ms, SSD read ~1 ms vs memory ~10 µs; Cash App sees ~3 minute savings for West Coast devs while far-flung developers (Australia) can be slower without edge replicas. System supports Gradle and Bazel.
AI-written summary · about
chapters (10)
- 0:00Introduction and setup
- 1:55What is a build cache
- 5:45Latency bottlenecks explained
- 7:08Geographic distance impact
- 9:20CDN solution for global cache
- 11:26Data center latency advantage
- 12:49Storage read speed differences
- 13:52Hot storage and compression
- 15:18Network protocol optimization
- 17:09Summary and free tier offer
related talks
38mThe power of cache in a slow world
27mImproving the Usability of Remote Caching
01h 04mCommon-sense acceleration of your MLOC build
29mLove your cache: Optimize for the second load
41mSpeeding up Jenkins, Gradle, and Maven with a Build Cache
30mThe Bitter and the Sweet of Running a Planet-Scale Build &...
23mWalk, Run, Fly - Take your build pipeline to the next level!
28mUniversal Cache: It’s time for fast, reliable, and cheap CI pipelines