Performance Story: How to Speed Up HTTP/2 Client
summary
Benchmarks use a simple JMH-style client/server loop and custom load tests and start with 25 req/s baseline on localhost; a TCP delay/TCP_NODELAY fix removes a OS networking stall and yields a large immediate boost. Profiling and a stack-trace sampling script finds excessive per-frame window updates; batching window updates and collapsing frame-to-bytebuffer writes gives ~23% throughput gain and turns a 10k/s plateau into 25k+/s scalability. Byte-buffer pools, direct vs heap buffers, and simplified no-pool code are measured; no-pool wins or ties and final performance matches Jetty-level “reasonable” targets.
AI-written summary · about
chapters (6)
related talks
40mWhirlwind tour through the HTTP2 spec
43mHTTP 2.0: why and how
01h 00mHTTP/2 : What you need to know
26mCan we Double the Node.js HTTP Client Throughput?
30mWhirlwind Tour through the HTTP2 Spec
48mHTTP 2
39mNetty 5: Lessons Learned
26mTuning Your Cloud: Improve Global Network Performance for Applications