Thread pools: variety of algorithms and features
summary
Compares concrete runtimes showing what actually causes slowdowns and how to fix them: measures with 500000 short tasks reveal long delays and queue contention when the default minimum threads equals processor count (12) and .NET threadpool’s hill climbing growth triggers spikes; explains the work stealing model used in many runtimes and the hill climbing plus Fourier based heuristics in .NET; builds a custom downloader thread pool with an external strategy, per‑thread init/teardown, throughput history and tail time prediction and shows it reduces contention and improves throughput.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:26Thread pool goals
- 5:03Bandwidth control and load balancing
- 10:02Thread pool types by platform
- 15:27Thread pool practical use
- 20:52Thread pool performance measurement
- 29:06Analyzing Netty's thread pool
- 31:42Fourier transform with thread pools
- 38:51Thread pool growth algorithms
- 44:51Thread pool platform comparison
- 45:31Building a custom thread pool
- 56:16Summary and optimization outlook
- 1:00:33Conclusion and Q&A
related talks
01h 00mMultithreading in .NET: When Performance Isn't Enough
01h 03mBuilding responsive and scalable applications
51mFrom Linux Threads To ForkJoinPool
11mImproving App Performance With Custom Thread Pools
01h 04mThreadPool for a Service That Adapts to External Load
01h 04mMaking fun with parallelism in .NET
04mInterview
30mThe untapped power of parallel programming: build efficient code that scales