Performance and Scale @ Istio Service Mesh
summary
Working group measures Istio end-to-end with Fortio for synthetic load and an open-source Blueproof Bookinfo-style app for realistic tests, automates nightly regression patrol across GCE, AWS, Azure, IBM Cloud and on-prem, and surfaces concrete fixes: remove excessive per-check logging, fix a Zipkin goroutine leak, eliminate duplicate Mixer calls and a bad cache-key (source IP). Results show p50 34→14ms and QPS 700→1700 after optimizations; goal is <10% mesh overhead.
AI-written summary · about
