A reworked TCP zero-copy receive API [LWN.net]
LWN.net needs you!Without subscribers, LWN would simply not exist. Please consider signing up for a subscription and helping to keep LWN publishing In April, LWN looked at the new API for zero-copy reception of TCP data that had been merged into the net-next tree for the 4.18 development cycle. After that article was written, a couple of issues came to the fore that required some changes to the AP
TIME_WAITã«é–¢ã™ã‚‹è©±
3. 自己紹介 - ã¾ãã¾ã MySQL ã§ã”飯食ã¹ã¦ã¾ã™ - 一時期㯠Resource Monitoring ã‚„ KVS ã«ã‚‚ 力入れã¦ã¾ã—㟠- ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯çš„ã«ã¯ç´ 人ã§ã™ - Linuxã¨ãƒãƒ¼ãƒ‰ã‚¦ã‚§ã‚¢ã¯å—œã‚€ç¨‹åº¦ - disk I/O ã«ã¯ã‚€ã‹ã—ã‹ã‚‰èˆˆå‘³ã‚ã‚Šã¾ã™ - ãã®ä»– slideshare ã¯ã“ã¡ã‚‰ - http://www.slideshare.net/takanorisejima/ 4. 本日ã®ãŠé¡Œ - kernel æ–°ã—ãã—ãŸã‚Šã™ã‚‹ã¨ã€TCPçš„ã«æ„è˜ã—ãŸã» ã†ãŒè‰¯ã„変化ãŒè¦‹ã¤ã‹ã‚‹ã®ã§ - 今日ã¯ã€Webアプリケーションサーãƒã®è¦³ç‚¹ã‹ ら〠connect(2) ã™ã‚‹éš›ã«æ°—ã«ãªã‚‹ TIME_WAIT ã«ã¤ã„ã¦ã€æ›¸ã„ã¦ã¿ã‚ˆã†ã‹ã¨æ€ã„ã¾ã™ - 有è˜è€…ã‹ã‚‰ã®ãƒžã‚µã‚«ãƒªã‚’ã€å¼·ãæ“è¿Žã„ãŸã—ã¾ã™ 5. 最åˆã«å‚考資料 - ã“ã®äºŒã¤ã®è¨˜äº‹ã‚’èªã‚“ã§ã„ãŸã ã‘ã°ã€ãã‚Œã§æ¦‚ ã良ã„ã¨æ€ã†ã‚“ã§ã™ãŒ
Why do we use the Linux kernel's TCP stack?
June 30, 2016 I’m at PolyConf in Poland today, and I watched this super interesting talk by Leandro Pereira about Lwan, an ~8000 line of code web server. He talked about a bunch of the optimizations they’d done (improve CPU cache performance! be really careful about locking!). You can read more about the performance on the website & the links there. It’s a super cool project because it started out
TCP small queues [LWN.net]
The "bufferbloat" problem is the result of excessive buffering in the network stack; it leads to long latencies and poor reliability in the network as a whole. Fixing it is a matter of buffering less data in each system between any two endpoints—a task that sounds simple, but proves to be more challenging than one might expect. It turns out that buffering can show up in many surprising places in t
Userspace networking
Seven years ago at LCA, Van Jacobsen introduced the concept of net channels but since then the concept of user mode networking has not hit the mainstream. There are several different user mode networking environments: Intel DPDK, BSD netmap, and Solarflare OpenOnload. Each of these provides higher performance than standard Linux kernel networking; but also creates new problems. This talk will expl
Why protocol stacks should be in user-space?
Why protocol stacks should be in user-space? Michio Honda (NEC Europe), JoaoTaveira Araujo (UCL), Luigi Rizzo (Universitea de Pisa), Costin Raiciu (Universitea Buchalest) Felipe Huici (NEC Europe) IETF87 MPTCP WG July 30, 2013 Berlin Motivation • Extending layer 4 functionality addresses a lot of problems – Increased performance • MPTCP, WindowScale, FastOpen,TLP, PRR – Ubiquitous encryption • Tcp
TCP Keepalive HOWTO
This document describes the TCP keepalive implementation in the linux kernel, introduces the overall concept and points to both system configuration and software development. Table of Contents1. Introduction1.1. Copyright and License1.2. Disclaimer1.3. Credits / Contributors1.4. Feedback1.5. Translations2. TCP keepalive overview2.1. What is TCP keepalive?2.2. Why use TCP keepalive?2.3. Checking fo
User-space Network Processing
3. • B3 k6 z 2 l • mTCP +4Intel4DPDK wi • github mTCP+4DPDK orz • Key4Value4Store w k Linux l • RADIS → • d v orz • Memcached → • d 3 4. A G 7 LNPMXXT 4 0 0.2 0.4 0.6 0.8 1 0 20 40 60 80 100 Key size (bytes) Key size CDF by appearance USR APP ETC VAR SYS 0 0.2 0.4 0.6 0.8 1 1 10 100 1000 10000 100000 1e+06 Value size (bytes) Value Size CDF by appearance USR APP ETC VAR SYS 0 0.2 0.4 0.6 0.8 1 1 10
tcp/dccp: lockless listener [LWN.net]
netdev <net[email protected]>, Eric Dumazet <edumazet@google.com>, Ying Cai <ycai@google.com>, Willem de Bruijn <willemb@google.com>, Eric Dumazet <eric.dumazet@gmail.com> TCP listener refactoring : this is becoming interesting ! This patch series takes the steps to use normal TCP/DCCP ehash table to store SYN_RECV requests, instead of the private per-listener hash table we had until now. SYNACK
How TCP backlog works in Linux
When an application puts a socket into LISTEN state using the listen syscall, it needs to specify a backlog for that socket. The backlog is usually described as the limit for the queue of incoming connections. Because of the 3-way handshake used by TCP, an incoming connection goes through an intermediate state SYN RECEIVED before it reaches the ESTABLISHED state and can be returned by the accept s
TCP keepaliveè¨å®šã§ãƒãƒ¼ãƒ•ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ã‚’解消
ï¼’å°ã®ã‚µãƒ¼ãƒé–“ãŒTCPãƒãƒ¼ãƒ•ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³çŠ¶æ…‹ã«ãªã‚‹äº‹è±¡ãŒç™ºç”Ÿã—ãŸã€‚ TCP keepalive(ã‚ープアライブ)ã®è¨å®šã§ãã®çŠ¶æ…‹ã‚’解消ã™ã‚‹ãŸã‚ã®æ‰‹æ®µã‚’メモã—ã¦ãŠã。 â—† æ§‹æˆ ã‚µãƒ¼ãƒAã¨ã‚µãƒ¼ãƒB â—† アプリケーション仕様 サーãƒAãŒã‚µãƒ¼ãƒB上ã®ãƒ•ã‚¡ã‚¤ãƒ«ã‚’定期的ã«å–å¾—ã™ã‚‹ã€‚ 両サーãƒé–“ã¯TCPコãƒã‚¯ã‚·ãƒ§ãƒ³ã‚’張り続ã‘ã¦ã„る。通信ã”ã¨ã«æŽ¥ç¶šã™ã‚‹ä»•æ§˜ã§ã¯ãªã„。 â—† トラブル サーãƒBã«éšœå®³ãŒç™ºç”Ÿã—ã€ã‚µãƒ¼ãƒBãŒçªç„¶ãƒ€ã‚¦ãƒ³ã—ãŸã€‚ ãã®ãŸã‚サーãƒBã‹ã‚‰ã‚µãƒ¼ãƒAã¸ã®TCPã®ã‚¯ãƒãƒ¼ã‚ºå‡¦ç†ãŒè¡Œã‚ã‚Œãšã€ サーãƒAã ã‘ã«TCPã®çŠ¶æ…‹ãŒESTABLISHEDã§æ®‹ã‚‹ãƒãƒ¼ãƒ•ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ã¨ãªã£ã¦ã—ã¾ã£ãŸã€‚ LANケーブルãŒæŠœã‘ãŸå ´åˆãªã©ã‚‚åŒã˜çŠ¶æ…‹ã«ãªã‚‹ã ã‚ã†ã‹ã€‚ ãƒãƒ¼ãƒ•ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ã®æœ‰ç„¡ã¯ã€ã‚µãƒ¼ãƒAã€ã‚µãƒ¼ãƒBã§ãã‚Œãžã‚Œ "lsof -n"ã€"netstat -an" ãªã©ã®ã‚³ãƒžãƒ³ãƒ‰ã‚’打ã¡ã€çŠ¶æ…‹ãŒESTABLISHEDã«ãª
conntrack ãŒãƒãƒ¼ãƒ‰ã•ã‚ŒãŸã‚µãƒ¼ãƒã§ LVS-DR を構æˆã™ã‚‹å ´åˆã®æ³¨æ„点 - Cybozu Inside Out | サイボウズエンジニアã®ãƒ–ãƒã‚°
ã“ã‚“ã«ã¡ã¯ã€‚インフラ開発ãƒãƒ¼ãƒ ã®æ·±è°·ã§ã™ã€‚ Linux ã® conntrack モジュールãŒãƒãƒ¼ãƒ‰ã•ã‚ŒãŸã‚µãƒ¼ãƒãƒ¼ã§ã€ãƒ¬ã‚¤ãƒ¤ãƒ¼4 (L4) ãƒãƒ¼ãƒ‰ãƒãƒ©ãƒ³ã‚µã‚’ LVS-DR ã§æ§‹æˆã—ãŸã¨ã“ã‚〠パケットフィルタãŒæ£å¸¸ã«åƒã‹ãšã€é€šä¿¡ã¯ã§ãã‚‹ã‚‚ã®ã®ã€ä¸€éƒ¨ã®ãƒ‘ケットã¯è½ã¡ã‚‹ã¨ã„ã†æŒ™å‹•ã«æ‚©ã¾ã•ã‚Œã¾ã—ãŸã€‚ 今回ã¯ã“ã®å•é¡Œã«ã¤ã„ã¦è©³ç´°ã«è§£æžã—ã€åŽŸå› ã¨å¯¾ç–ãŒã‚ã‹ã£ãŸã®ã§ã”紹介ã„ãŸã—ã¾ã™ã€‚ 発生ã—ãŸå•é¡Œ 今回ã®å•é¡Œã¯ã€Ubuntu を使ã£ã¦ L4 ãƒãƒ¼ãƒ‰ãƒãƒ©ãƒ³ã‚µã‚’実装ã—ã¦ã„る時ã«éé‡ã—ã¾ã—ãŸã€‚ 構築ã—ãŸãƒãƒ¼ãƒ‰ãƒãƒ©ãƒ³ã‚µã«ã¯ã€ã‚»ã‚ュリティ対ç–ã®ãŸã‚ iptables ã®ãƒ©ãƒƒãƒ‘ーã§ã‚ã‚‹ UFW を使ã£ã¦ã€ パケットフィルタをã‹ã‘ã¦ã„ã¾ã—ãŸã€‚ ãƒãƒ¼ãƒ‰ãƒãƒ©ãƒ³ã‚µã®å®Ÿè£…ã«ã¯ã€Linux カーãƒãƒ«ã«å®Ÿè£…ã•ã‚Œã¦ã„ã‚‹ LVS を使ã„ã¾ã—ãŸã€‚ LVS ã§ã¯ã„ãã¤ã‹ã®ãƒ‘ケット転é€æ–¹å¼ã‚’サãƒãƒ¼ãƒˆã—ã¦ãŠã‚Šã€ä»Šå›žã¯ãƒ‘フォーマンスã«å„ªã‚Œã‚‹
The SO_REUSEPORT socket option [LWN.net]
One of the features merged in the 3.9 development cycle was TCP and UDP support for the SO_REUSEPORT socket option; that support was implemented in a series of patches by Tom Herbert. The new socket option allows multiple sockets on the same host to bind to the same port, and is intended to improve the performance of multithreaded network server applications running on top of multicore systems. Th
mTCP使ã£ã¦ã¿ãŸ
mTCP enables high-performance userspace TCP/IP stacks by bypassing the kernel and reducing system call overhead. It was shown to achieve up to 25x higher throughput than Linux for short flows. The document discusses porting the iperf benchmark to use mTCP, which required only minor changes. Performance tests found that mTCP-ified iperf achieved similar throughput as Linux iperf for different packe
listen()ã®backlogãŒä¸è¶³ã—ãŸéš›ã®TCP_DEFER_ACCEPTã®å‹•ä½œã«ã¤ã„㦠- blog.nomadscafe.jp
TCP_DEFER_ACCEPTã¯ã€Linuxã§ã‚µãƒãƒ¼ãƒˆã•ã‚Œã¦ã„ã‚‹TCPã®ã‚ªãƒ—ションã§ã€ã‚µãƒ¼ãƒå´ã§ä½¿ç”¨ã—ãŸå ´åˆã«ã¯accept(2)ã‹ã‚‰ã®ãƒ–ãƒãƒƒã‚¯è§£é™¤ã‚’TCP接続ãŒå®Œäº†ã—ãŸã‚¿ã‚¤ãƒŸãƒ³ã‚°ã§ã¯ãªã最åˆã®ãƒ‡ãƒ¼ã‚¿ãŒåˆ°ç€ã—ãŸã‚¿ã‚¤ãƒŸãƒ³ã‚°ã§è¡Œã£ã¦ãれるオプションã§ã™ã€‚ Webサーãƒãƒ»ã‚¢ãƒ—リケーションサーãƒã§ã¯ãƒªã‚¯ã‚¨ã‚¹ãƒˆãŒåˆ°ç€ã—ã¦ã‹ã‚‰accept(2)ã®ãƒ–ãƒãƒƒã‚¯ã‚’解除ã™ã‚‹ã®ã§ã€ãƒªã‚¯ã‚¨ã‚¹ãƒˆã®åˆ°ç€ã‚’Webサーãƒãƒ»ã‚¢ãƒ—リケーションサーãƒã§å¾…ã¤å¿…è¦ãŒãªããªã‚Šã€ç‰¹ã«preforkåž‹ã®ã‚µãƒ¼ãƒã§ã¯åŠ¹çŽ‡çš„ã«ãƒ—ãƒã‚»ã‚¹ã‚’使ãˆã‚‹ã‚ˆã†ã«ãªã‚‹ã¨ã„ã†åˆ©ç‚¹ãŒã‚ã‚Šã¾ã™ã€‚Perlã§ã¯StarletãŒã“ã®æ©Ÿèƒ½ã‚’有効ã«ã—ã¦ã„ã¾ã™ ã¨ã“ã‚ãŒã€æŸã‚µãƒ¼ãƒ“スã§TCP_DEFER_ACCEPTãŒæœ‰åŠ¹ã«ã‚‚é–¢ã‚らãšã€accept後ã®readã§ãƒ‡ãƒ¼ã‚¿ãŒèªã‚ãšã€æœ€æ‚ªã®å ´åˆã€ãƒ‡ãƒ•ã‚©ãƒ«ãƒˆã®timeoutã§ã‚ã‚‹5分間プãƒã‚»ã‚¹ãŒã‚¹ãƒˆãƒ¼ãƒ«ã™ã‚‹ã“ã¨ãŒã‚ã‚Šã¾ã—ãŸã€‚strace
Linuxã§TCP_DEFER_ACCEPTãŒæœ‰åŠ¹ã§ã‚‚accept後readã§ããªã„ç†ç”±
listen()ã®backlogãŒä¸è¶³ã—ãŸéš›ã®TCP_DEFER_ACCEPTã®å‹•ä½œã«ã¤ã„㦠- blog.nomadscafe.jpã¨ã„ã†è¨˜äº‹ã®ä¸ã§ã€listen backlog ãŒã‚ãµã‚ŒãŸå¾Œã« accept(2) ã™ã‚‹ã¨ã€ãã®å¾Œã® read(2) ㌠EAGAIN ã‚’è¿”ã—ãŸã‚Šã€æŽ¥ç¶šãŒä¸å®‰å®šã«ãªã‚‹ã¨ã„ã†äº‹è±¡ãŒèª¬æ˜Žã•ã‚Œã¦ã„ã¾ã—ãŸã€‚æ°—ã«ãªã£ãŸã®ã§èª¿ã¹ã¦ã¿ãŸã“ã¨ã‚’ã¾ã¨ã‚ã¾ã™ã€‚ çµè«–ã‹ã‚‰è¨€ã†ã¨ã“ã‚Œã¯Linuxã®ä»•æ§˜ã§ã™ã€‚manã®tcp(7)を見るã¨ã€ TCP_DEFER_ACCEPT (since Linux 2.4) Allow a listener to be awakened only when data arrives on the socket. Takes an integer value (seconds), this can bound the maximum number of
「net.ipv4.tcp_tw_recycleã€ã‚’有効ã«ã™ã‚‹ã®ã¯ï¼ˆå ´åˆã«ã‚ˆã£ã¦ã¯ï¼‰ã‚„ã‚ãŸæ–¹ãŒã„ã„ - pullphone's blog
ãã‚‚ãも「tcp_tw_recycleã€ã£ã¦ãªã«ï¼Ÿ TIME_WAIT状態ã®ã‚½ã‚±ãƒƒãƒˆ*1を高速ã«å†åˆ©ç”¨ã™ã‚‹ãŸã‚ã®Linuxカーãƒãƒ«ç‰¹æœ‰ã®ä»•çµ„ã¿ã‚‰ã—ã„。 「/etc/sysctl.confã€ã§ã“ã„ã¤ï¼ˆnet.ipv4.tcp_tw_recycle）を1ã«ã—ã¦ã‚„ã£ã¦ã€Œsysctl -pã€ã™ã‚‹ã ã‘ã§æœ‰åŠ¹ã«ãªã‚Šã¾ã™ã€‚ 「TIME_WAITã®ã‚½ã‚±ãƒƒãƒˆã‚’å°‘ãªãã—ã¦ãれるんã§ã—ょ？ã„ã„ã˜ã‚ƒã‚“ï¼ã€ã¨ã„ã†æ„Ÿã˜ã§è¨å®šã—ã¦ã—ã¾ã„ãã†ã«ãªã‚Šã¾ã™ãŒãƒ»ãƒ»ãƒ» ãªã‚“ã§ãƒ€ãƒ¡ãªã®ï¼Ÿ çµè«–ã‹ã‚‰è¨€ã†ã¨ã€ã“ã„ã¤ã‚’有効ã«ã—ãŸã¨ãã€åŒã˜ã‚°ãƒãƒ¼ãƒãƒ«IPã®ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã‹ã‚‰ã®æŽ¥ç¶šã§ã‹ã¤TCPパケットã«ã‚¿ã‚¤ãƒ ã‚¹ã‚¿ãƒ³ãƒ—æƒ…å ±ãŒå…¥ã£ã¦ã„ã‚‹å ´åˆã§ã€ã»ã¼åŒæ™‚ã«ãƒ‘ケットをé€ã‚‹ã¨ã€å¤ã„タイムスタンプã®æ–¹ã®ãƒ‘ケットをå‹æ‰‹ã«ãƒ‰ãƒãƒƒãƒ—ã—ã¡ã‚ƒã„ã¾ã™ã€‚ ã©ã†ã„ã†ã“ã¨ãªã®ãƒ»ãƒ»ãƒ» 「tcp_tw_recycleã€ã¯ã€ã€ŒåŒä¸€IPã‹ã‚‰ã®ãƒ‘ケットãŒåˆ°ç€ã—ãŸã¨ãã€ä½¿ã£ã¦ã„ãŸã‚½ã‚±
Linuxã®TCP SYNã®å†é€é–“éš”ã®åˆæœŸå€¤ãŒ3秒ã‹ã‚‰1秒ã«å¤‰æ›´ã•ã‚Œã¦ã„㟠- å…ƒRX-7ä¹—ã‚Šã®é©å½“ãªæ—¥ã€…
ã¨ã„ã†ã“ã¨ã«ã€(今更？)気付ã„ãŸãŠè©±ã§ã™ã€‚ HAを組んã éš›ã®VIPã®åˆ‡ã‚Šæ›¿ãˆãƒ†ã‚¹ãƒˆã‚’ã‚„ã£ã¦ã„ã‚‹ã¨ãã«ã€é«˜è² è·æ™‚ã¨ã‹ã¯åˆ‡ã‚Šæ›¿ãˆã«7秒ã´ã£ãŸã‚Šã‹ã‹ã‚‹ã‚±ãƒ¼ã‚¹ã¨ã‹ãŒã‚ã£ã¦ã€7秒ã£ã¦ä½•ã®æ•°å—ã ã‚ã†ã¨ç–‘å•ã‚’æŒã¡ã¾ã—ãŸã€‚ OSã¯ã€CentOS 6.4(2.6.32-358.23.2.el6.x86_64)ã§ã™ã€‚ TCP SYNã®å†é€é–“éš”ãŒã€1...2...4...秒ã«ãªã£ã¦ã„ã‚‹ ã§ã€tcpdumpを眺ã‚ã¦ã„ã‚‹ã¨ä»¥ä¸‹ã®ã‚ˆã†ãªã‚·ãƒ¼ã‚±ãƒ³ã‚¹ã§ã™ã€‚ 11:50:35.689301 IP client-host.8957 > server-host.http: Flags [S], seq 1616681830, win 14600, options [mss 1460,sackOK,TS val 889880946 ecr 0,nop,wscale 7], length 0 11:50:36.688503 IP
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
https://www.kernel.org/doc/Documentation/networking/proc_net_tcp.txt
hiboma/kernel/net/net-backlog.md at master · hiboma/hiboma
{{#tags}}- {{label}}
{{/tags}}