NUMA-aware scheduler for Go
NUMA-aware scheduler for Go Dmitry Vyukov, dvyukov@ Sep 2014 AbstractWith the advent of Intel QPIÂ and AMD HTÂ technologies even commodity SMP servers become NUMA. As core counts continue to grow, this trend will continue as well. When per-package computational power will reach a threshold, even a single package will require multiple memory lanes effectively becoming NUMA (see e.g. Tile-Gx72Â process
DMM inside
ãªã›ã‚™DMMã‹ã‚™web3ã«å‚å…¥ã—ãŸã®ã‹ã€‚Seamoon Protocolã‹ã‚™ç›®æŒ‡ã™æ–°ãŸãªã‚¨ãƒ³ã‚¿ãƒ¡ä½“験ã®æœªæ¥ã¨ã¯
NUMAアーã‚テクãƒãƒ£ã¨ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯æ€§èƒ½ï¼ˆï¼‘）ï¼èª¬æ˜Žç·¨ - Blog
NUMA(Non-Uniform Memory Access)ã¨ã¯ Non-uniform memory access - Wikipedia, the free encyclopedia NUMA - Wikipedia NUMA（Non-Uniform Memory Accessã€ãƒŒãƒžï¼‰ã¨ã¯ã€å…±æœ‰ãƒ¡ãƒ¢ãƒªåž‹ãƒžãƒ«ãƒãƒ—ãƒã‚»ãƒƒã‚µã‚³ãƒ³ãƒ”ュータシステムã®ã‚¢ãƒ¼ã‚テクãƒãƒ£ã®ã²ã¨ã¤ã§ã€è¤‡æ•°ãƒ—ãƒã‚»ãƒƒã‚µãŒå…±æœ‰ã™ã‚‹ãƒ¡ã‚¤ãƒ³ãƒ¡ãƒ¢ãƒªã¸ã®ã‚¢ã‚¯ã‚»ã‚¹ã‚³ã‚¹ãƒˆãŒã€ãƒ¡ãƒ¢ãƒªé ˜åŸŸã¨ãƒ—ãƒã‚»ãƒƒã‚µã«ä¾å˜ã—ã¦å‡ä¸€ã§ãªã„アーã‚テクãƒãƒ£ã§ã‚る。 複数CPUãŒã‚ã‚‹å ´åˆã€è¿‘ã„メモリã¨é ã„メモリãŒæ§‹æˆä¸Šå˜åœ¨ã—ã€ãƒ‡ãƒ¼ã‚¿ãŒç½®ã‹ã‚Œã‚‹å ´æ‰€ã«ã‚ˆã£ã¦ã€æ€§èƒ½ã«å·®ãŒå‡ºã¦ã—ã¾ã†ã‚¢ãƒ¼ã‚テクãƒãƒ£ã§ã‚る。 ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯æ€§èƒ½ã‚’測定ã™ã‚‹å ´åˆã‚‚ã€NUMA構æˆä¸Šä¸åˆ©ãªçŠ¶æ…‹ã§æ€§èƒ½æ¸¬å®šã‚’ã—ã¦ã—ã¾ã†ã¨ã€æ€ã†ã‚ˆã†ãªæ€§èƒ½ãŒå‡ºãªã„å ´åˆãŒã‚ã‚‹ãŸã‚ã€æ³¨æ„ãŒå¿…è¦ã§ã‚る。本稿ã§ã¯ã€NUMAåŠã³CPU
Optimizing Linux Memory Management for Low-latency / High-throughput Databases
Optimizing Linux Memory Management for Low-latency / High-throughput Databases Co-author: Cuong Tran Table of Contents Introduction Setting up the context Reproducing and understanding Linux's zone reclaim behavior NUMA memory rebalancing also triggers direct page scans Lessons learned Introduction GraphDB is the storage layer of LinkedIn's real-time distributed social graph service. Our service h
MongoDBã‚’NUMAãªãƒžã‚·ãƒ³ã§ä½¿ã†ã¨ãã®æ³¨æ„ - 酒日記 ã¯ã¦ãªæ”¯åº—
デュアルCPUã§è¨ˆ12コア24スレッドã€ãƒ¡ãƒ¢ãƒª48GBã¨ã„ã†ãƒžã‚·ãƒ³ã§ MongoDB-2.0.8 ã‚’ã—ã°ã‚‰ã稼åƒã•ã›ãŸã¨ã“ã‚ã€çªç„¶ CPU ã® system time ãŒ1コア分暴走ã—ãŸã‚ˆã†ã«ãªã‚‹ã€ã¨ã„ã†ç¾è±¡ãŒèµ·ãã¾ã—ãŸã€‚ 最åˆã¯åŽŸå› ãŒã‚ˆã分ã‹ã‚‰ãšã€ã¨ã‚Šã‚ãˆãš mongod ã®ãƒ—ãƒã‚»ã‚¹ã‚’ kill ã—ã¦èµ·å‹•ã—ç›´ã—ãŸã‚‰å¾©æ—§ã—ãŸã®ã§ã™ãŒã€ã¾ãŸã—ã°ã‚‰ãã™ã‚‹ã¨åŒã˜ç¾è±¡ã«ã€‚ mongoã®ãƒ¡ãƒ¢ãƒªä½¿ç”¨é‡ã¨ Load Average をプãƒãƒƒãƒˆã—ã¦ã¿ã‚‹ã¨ã€ã©ã†ã‚‚ã‚る程度 (ç´„24GB?) ã®ãƒ¡ãƒ¢ãƒªã‚’使ã£ãŸã¨ã“ã‚ã§æš´èµ°ãŒèµ·ãã¦ã„るよã†ãªâ€¦â€¦ã¨ãƒã‚°ã‚’見直ã—ã¦ã¿ã‚‹ã¨ã€èµ·å‹•æ™‚ã« WARNING ãŒã§ã¦ã„ã¾ã—ãŸã€‚ Sun Jan 20 00:10:01 [initandlisten] MongoDB starting : pid=12669 port=27017 dbpath=/var/lib/mongo 64-b
1
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
Thread and Memory Placement on NUMA Systems: Asymmetry Matters | USENIX
The MySQL “swap insanity†problem and the effects of the NUMA architecture
Automatic NUMA Balancing
lanceshelton/irqstat · GitHub - A better way to watch /proc/interrupts, designed for NUMA systems with many processors.
{{#tags}}- {{label}}
{{/tags}}