[B! c10k] ttakezawaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

ttakezawa id:ttakezawa

c10kã«é–¢ã™ã‚‹ttakezawaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (6)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

Monoceros ã¨ã„ã†Preforkåž‹ã ã‘ã©C10Kã®æŽ¥ç¶šã‚’æŒãã“ã¨ãŒã§ãã‚‹PSGI/Plackã‚µãƒ¼ãƒæ›¸ãã¾ã—ãŸ - blog.nomadscafe.jp
Monoceros ã¨ã„ã†PSGI/Plackã‚µãƒ¼ãƒæ›¸ãã¾ã—ãŸ https://metacpan.org/release/Monoceros https://github.com/kazeburo/Monoceros Starmanã‚„Starletã®ã‚ˆã†ãªPreforkãªã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã‚µãƒ¼ãƒã§ã¯ã€ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ã®ç¶æŒã‚¤ã‚³ãƒ¼ãƒ«ãƒ—ãƒã‚»ã‚¹ã®å æœ‰ãªã®ã§ã€HTTPã®KeepAliveã¯ç„¡åŠ¹ã«ã™ã‚‹ã®ãŒä¸€èˆ¬çš„ã§ã™ãŒã€è² è·ã®é«˜ã„ã‚µãƒ¼ãƒ“ã‚¹ã§ã¯TIME_WAITçŠ¶æ…‹ã®ã‚½ã‚±ãƒƒãƒˆãŒæºœã¾ã£ãŸã‚Šã€SYN-ACKã®å†é€å•é¡Œãªã©ã‚ã‚Šã€KeepAliveã‚’ä½¿ã„ãŸã„ã¨ã„ã†æ¬²æ±‚ãŒã‚ã£ãŸã‚Šãªã‹ã£ãŸã‚Šã—ã¾ã™ã€‚ Monoceros ã¯ãƒªã‚¯ã‚¨ã‚¹ãƒˆã‚’å‡¦ç†ã™ã‚‹workerã®ä»–ã«ã€ã‚¤ãƒ™ãƒ³ãƒˆãƒ‰ãƒªãƒ–ãƒ³ã§å‹•ãã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ç®¡ç†ãƒ—ãƒã‚»ã‚¹ã‚’ç«‹ã¦ã¦ã€ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã‹ã‚‰ã®æŽ¥ç¶šã‚½ã‚±ãƒƒãƒˆã‚’unix domain socketã‚’ä½¿ã„ãƒ—ãƒã‚»ã‚¹é–“ã§ã‚„ã‚Šã¨ã‚Šã—ã¾ã™ã€‚å¾…æ©Ÿ
ttakezawa 2013/10/01
server

plack

monoceros

performance

benchmark

keepalive

c10k
ãƒªãƒ³ã‚¯
node.js ã¨ã¯ä½•ã‹ (2) - I am bad at math
æ˜¨æ—¥ã«å¼•ãç¶šãã€ã„ã–ï¼part2ãªã®ã ã€‚ å‰å›žã§ã¯ node.js ã¨ v8 ã®çµã³ã¤ãã¾ã§ã‚’æ›¸ã„ãŸã®ã§ã€ä»Šæ—¥ã¯ Non-Blocking I/O ã®è©±ã‚’ã€‚ Non-Blocking I/O ã¨ã„ã†è¨€è‘‰ã‹ã‚‰ãƒ–ãƒãƒƒã‚¯ã—ãªã„ I/O ã‚’ã‚¤ãƒ¡ãƒ¼ã‚¸ã™ã‚‹ã®ã¯ãŸã‚„ã™ã„ã€‚ã§ã‚‚ã“ã‚Œã‚’å®Ÿç¾ã—ã‚ˆã†ã¨ãªã‚‹ã¨ã„ã‚ã„ã‚ã¨ã¾ã‚é¢å€’ãã•ã„ã‚“ã ã‚ˆãã€‚ ãã‚Œã‚’è§£æ±ºã™ã‚‹å¸¸å¥—æ‰‹æ®µã§è¨€ã†ã¨ãƒ•ã‚¡ã‚¤ãƒ«ãƒ‡ã‚£ã‚¹ã‚¯ãƒªãƒ—ã‚¿ï¼ˆãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ãªã‚‰ã‚½ã‚±ãƒƒãƒˆã ãï¼‰ã‚’é–‹ã„ã¦ãã‚Œã‚’selectã‚·ã‚¹ãƒ†ãƒ ã‚³ãƒ¼ãƒ«ã®ç›£è¦–å¯¾è±¡ã«åŠ ãˆã¦ãŠãã€selectã‚’å‘¼ã³å‡ºã™ã“ã¨ã§ç›£è¦–ã™ã‚‹ã£ã¦ã„ã†æ–¹æ³•ãŒã‚ã‚‹ã€‚ã“ã†ã™ã‚‹ã¨ä½•ãŒå¬‰ã—ã„ã®ã‹ãƒ•ã‚¡ã‚¤ãƒ«ãƒ‡ã‚£ã‚¹ã‚¯ãƒªãƒ—ã‚¿ãŒï¼’ã¤ã‚ã‚‹å ´åˆã§è€ƒãˆã¦ã¿ã‚ˆã†ã€‚ ã¾ãšAã¨Bã¨ã„ã†ãƒ•ã‚¡ã‚¤ãƒ«ãƒ‡ã‚£ã‚¹ã‚¯ãƒªãƒ—ã‚¿ã‚’ç›£è¦–å¯¾è±¡ã¨ã™ã‚‹ã€‚ selectã‚·ã‚¹ãƒ†ãƒ ã‚³ãƒ¼ãƒ«ã‚’å‘¼ã³å‡ºã—ã€ãã®ã©ã¡ã‚‰ã‹ãŒèªã¿å‡ºã—æº–å‚™å®Œäº†ã¨ãªã£ã¦ã„ãªã„ã‹ã‚’ç¢ºèªã™ã‚‹ã€‚ ã‚‚ã—ã©ã£ã¡ã‚‚æº–å‚™ã§ãã¦ã„ãªã‹ã£ãŸã‚‰ãƒ—ãƒã‚»
ttakezawa 2010/10/22
network

library

c10k

conference
ãƒªãƒ³ã‚¯
Kazuho@Cybozu Labs: ã€Œã‚µãƒ¼ãƒæ›¸ããªã‚‰ epoll ä½¿ã†ã¹ãã€ã¯ã€ä»Šã§ã‚‚æ£ã—ã„ã®ã‹
å¤šæ•°ã®TCPæŽ¥ç¶šã‚’ãƒãƒ³ãƒ‰ãƒªãƒ³ã‚°ã™ã‚‹ã‚µãƒ¼ãƒã‚’æ›¸ããªã‚‰ã€1ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³1ã‚¹ãƒ¬ãƒƒãƒ‰ã®ãƒ¢ãƒ‡ãƒ«ã§ã¯ãªãã€epollã‚„kqueueã®ã‚ˆã†ãªã‚¤ãƒ™ãƒ³ãƒˆé§†å‹•åž‹ã®I/Oå¤šé‡åŒ–ã‚’è¡Œã†ã¹ãã ã€ã¨è¨€ã‚ã‚Œã¾ã™ã€‚ã ãŒã€ãã®ã‚ˆã†ãªä¸»å¼µã¯ã€ã€ŒC10Kå•é¡Œã€ãŒæ›¸ã‹ã‚ŒãŸ2002å¹´ã‹ã‚‰7å¹´çµŒéŽã—ãŸä»Šã§ã‚‚æœ‰åŠ¹ãªã®ã§ã—ã‚‡ã†ã‹ï¼Ÿã€€echoã‚µãƒ¼ãƒã‚’æ›¸ã„ã¦ã€ãƒ™ãƒ³ãƒãƒžãƒ¼ã‚¯ã‚’å–ã£ã¦ã¿ã‚‹ã“ã¨ã«ã—ã¾ã—ãŸã€‚ ãµãŸã¤ã®ã‚°ãƒ©ãƒ•ã¯ã€ã„ãšã‚Œã‚‚æŽ¥ç¶šæ•°ã¨ã‚¹ãƒ«ãƒ¼ãƒ—ãƒƒãƒˆã®é–¢ä¿‚ã‚’è¡¨ã—ã¦ã„ã¾ã™ã€‚æœ€åˆã®ã‚°ãƒ©ãƒ•ã¯ã€å…¨æŽ¥ç¶šãŒã‚¢ã‚¯ãƒ†ã‚£ãƒ–ã«é€šä¿¡ã—ãŸå ´åˆã€ã‚ã¨ã®ã‚°ãƒ©ãƒ•ã¯ã€å…¨æŽ¥ç¶šã®ã†ã¡å°æ•°ã®ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ãŒé †æ¬¡ã‚¢ã‚¯ãƒ†ã‚£ãƒ–ã«ãªã£ã¦ã„ãã€ã¨ã„ã†ãƒ¢ãƒ‡ãƒ«ã§ã™ã€‚ã“ã‚Œã‚‰ã®ã‚°ãƒ©ãƒ•ã‹ã‚‰ã€ä»¥ä¸‹ã‚ˆã†ãªã“ã¨ãŒèªã¿å–ã‚Œã¾ã™ã€‚ epoll ã‚‚ per-thread ãƒ¢ãƒ‡ãƒ«ã‚‚ã€è‰¯ãã‚¹ã‚±ãƒ¼ãƒ«ã™ã‚‹ epoll ã¯ã€ãƒ¯ãƒ¼ã‚¯ã‚»ãƒƒãƒˆãŒå°ã•ã„å ´åˆã« (æœ€å¤§50%) per-thread ãƒ¢ãƒ‡ãƒ«ã‚ˆã‚Šã‚‚é«˜é€Ÿ å°‘ãªãã¨ã‚‚ã€1ã‚³ãƒ
ttakezawa 2009/09/27
linux

c10k

concurrent

epoll

server

performance
ãƒªãƒ³ã‚¯
Kazuho@Cybozu Labs: Picoev: a tiny event loop for network applications, faster than libevent or libev
I am sure many programmers writing network applications have their own abstracting layers hiding the differences between various I/O multiplex APIs, like select(2), poll(2), epoll(2), ... And of course, I am one among them.Â While writing mycached (see Mycached: memcached protocol support for MySQL for more information), I was at first considering of using libev for multiplexing socket I/Os.Â Libe
ttakezawa 2009/09/10
c

library

c10k

libev

performance
ãƒªãƒ³ã‚¯
TheC10kProblem - ã€ŒC10Kå•é¡Œã€ï¼ˆã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆ1ä¸‡å°å•é¡Œï¼‰ã¨ã¯ã€ãƒãƒ¼ãƒ‰ã‚¦ã‚§ã‚¢ã®æ€§èƒ½ä¸Šã¯å•é¡ŒãŒãªãã¦ã‚‚ã€ã‚ã¾ã‚Šã«ã‚‚ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã®æ•°ãŒå¤šããªã‚‹ã¨ã‚µãƒ¼ãƒãŒãƒ‘ãƒ³ã‚¯ã™ã‚‹å•é¡Œã®ã“ã¨
TheC10kProbl em - ã€ŒC10Kå•é¡Œã€ï¼ˆã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆ1ä¸‡å°å•é¡Œï¼‰ã¨ã¯ã€ãƒãƒ¼ãƒ‰ã‚¦ã‚§ã‚¢ã®æ€§èƒ½ä¸Šã¯å•é¡ŒãŒãªãã¦ã‚‚ã€ã‚ã¾ã‚Šã«ã‚‚ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã®æ•°ãŒå¤šããªã‚‹ã¨ã‚µãƒ¼ãƒãŒãƒ‘ãƒ³ã‚¯ã™ã‚‹å•é¡Œã®ã“ã¨ ç›®æ¬¡ ã“ã®æ–‡æ›¸ã«ã¤ã„ã¦ C10K å•é¡Œ é–¢é€£ã‚µã‚¤ãƒˆ ã¾ãšèªã‚€ã¹ãæœ¬ I/O ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ I/O æˆ¦ç•¥ 1. å„ã‚¹ãƒ¬ãƒƒãƒ‰ãŒè¤‡æ•°ã®ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã‚’å—ã‘ä»˜ã‘ã‚‹. ãã—ã¦ãƒŽãƒ³ãƒ–ãƒãƒƒã‚ãƒ³ã‚° I/O ã¨ ãƒ¬ãƒ™ãƒ«ãƒ»ãƒˆãƒªã‚¬åž‹ã®å®Œäº†é€šçŸ¥ã‚’åˆ©ç”¨ã™ã‚‹. ä¼çµ±çš„ãª select() ä¼çµ±çš„ãª poll() /dev/poll kqueue() 2. å„ã‚¹ãƒ¬ãƒƒãƒ‰ãŒè¤‡æ•°ã®ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã‚’å—ã‘ä»˜ã‘ã‚‹. ãã—ã¦ãƒŽãƒ³ãƒ–ãƒãƒƒã‚ãƒ³ã‚° I/O ã¨ å¤‰æ›´åž‹ã®å®Œäº†é€šçŸ¥(readiness change notification)ã‚’åˆ©ç”¨ã™ã‚‹. kqueue() epoll ãƒªã‚¢ãƒ«ã‚¿ã‚¤ãƒ ãƒ»ã‚·ã‚°ãƒŠãƒ« fd å˜ä½ã®ã‚·ã‚°ãƒŠãƒ« (Signal-per-fd)
ttakezawa 2009/02/06
server

linux

performance

c10k
ãƒªãƒ³ã‚¯
libev v.s. libevent
ttakezawa 2008/12/29
c10k

server

performance

libev

library
ãƒªãƒ³ã‚¯
1