Optimizing web servers for high throughput and low latency
This is an expanded version of my talk at NginxConf 2017 on September 6, 2017. As an SRE on the Dropbox Traffic Team, I’m responsible for our Edge network: its reliability, performance, and efficiency. The Dropbox edge network is an nginx-based proxy tier designed to handle both latency-sensitive metadata transactions and high-throughput data transfers. In a system that is handling tens of gigabit
How to achieve low latency with 10Gbps Ethernet
Good morning! In a recent blog post we explained how to tweak a simple UDP application to maximize throughput. This time we are going to optimize our UDP application for latency. Fighting with latency is a great excuse to discuss modern features of multiqueue NICs. Some of the techniques covered here are also discussed in the scaling.txt kernel document. CC BY-SA 2.0 image by Xiaojun Deng Our expe
ã‚ãªãŸã®ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã‚¹ã‚¿ãƒƒã‚¯æ£ã—ãè¨å®šã•ã‚Œã¦ã„ã¾ã™ã‹ï¼Ÿ - Qiita
#ã¯ã˜ã‚ã« Linux Advent Calendar 10 日目ã®è¨˜äº‹ã§ã™ã€‚ é‹ç”¨ã‚„ç ”ç©¶é–‹ç™ºã®ç¾å ´ã§ã¯ã€ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢ã®å®Ÿé¨“ã€ã¾ãŸã¯æ©Ÿå™¨ã®ãƒ†ã‚¹ãƒˆã‚„é¸å®šãªã©ã®ãŸã‚ã«ã€ãƒ™ãƒ³ãƒãƒžãƒ¼ã‚¯ãƒ„ールや自å‰ã®ã‚¢ãƒ—リケーションã§ã‚³ãƒ³ãƒ”ュータ間ã®é€šä¿¡é€Ÿåº¦ã‚’計測ã™ã‚‹æ©Ÿä¼šãŒå¤šã€…ã‚ã‚‹ã¨æ€ã„ã¾ã™ã€‚一方ã§10Gbpsã‚„40Gbpsã¨ã„ã£ãŸæ˜¨ä»Šã®é«˜é€Ÿãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã«ãŠã„ã¦ã¯ã€ã“れらã®è¨ˆæ¸¬çµæžœã¯ã‚¢ãƒ—リケーションã®é€šä¿¡API部分ã®å®Ÿè£…ã€ã‚«ãƒ¼ãƒãƒ«ãƒ‘ラメータã¾ãŸã¯ã‚³ãƒ³ãƒ‘イルオプションã«ã‚ˆã£ã¦å¤§ãã変ã‚ã£ã¦ã—ã¾ã†ãŸã‚ã€æ£ç¢ºãªè¨ˆæ¸¬ã‚’è¡Œã†ãŸã‚ã«ã¯ã“れらをæ£ã—ãè¨å®š/ç†è§£ã™ã‚‹å¿…è¦ãŒã‚ã‚Šã¾ã™ã€‚ã“ã®è¨˜äº‹ã§ã¯ã€ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯å‘¨ã‚Šã®ã‚«ãƒ¼ãƒãƒ«ã¨ã‚¢ãƒ—リケーションã®å‹•ä½œã®æ¦‚è¦ã¨ã€ãã®ä¸ã®é‡è¦ãªãƒã‚¤ãƒ³ãƒˆã‚’ç†è§£ã™ã‚‹ã“ã¨ã‚’目的ã«ã—ã¾ã™ã€‚ #ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ãƒ—ãƒã‚°ãƒ©ãƒŸãƒ³ã‚°ã®ãŠã•ã‚‰ã„ ã¾ãšæœ€åˆã«ã€TCPを使ã†ä»Šæ™‚ã®ã‚µãƒ¼ãƒãƒ—ãƒã‚°ãƒ©ãƒ ãŒã©ã®ã‚ˆã†ã«ã§ãã¦ã„ã‚‹ã‹ç°¡å˜ã«ãŠã•ã‚‰ã„ã—
Linux Networking Explained
1. Linux Networking Explained LinuxCon 2016, Toronto Thomas Graf (@tgraf__) Kernel, Cilium & Open vSwitch Team Noiro Networks (Cisco) 2. Did you catch part I? â— Part II: LinuxCon, Toronto, 2016 Linux Networking Explained Network devices, Namespaces, Rou"ng, Veth, VLAN, IPVLAN, MACVLAN, MACVTAP, Bonding, Team, OVS, Bridge, BPF, IPSec â— Part I: LinuxCon, Sea,le, 2015 Kernel Networking Walkthrough Th
Netexpertise - Display Network Stats on Linux with Iptables
As for any firewall, iptables is able to do network statistics reporting. the -v (–verbose) option along with the list switch (-L) show packets and bytes counters. Network stats are available on a per rule basis. Here is an example on the INPUT chain: [stats@linux_server]$ sudo iptables -nvL INPUT Chain INPUT (policy DROP 74941 packets, 7900K bytes) pkts bytes target prot opt in out source destina
18 Commands to Monitor Network Bandwidth on Linux server
Network monitoring on Linux This post mentions some linux command line tools that can be used to monitor the network usage. These tools monitor the traffic flowing through network interfaces and measure the speed at which data is currently being transferred. Incoming and outgoing traffic is shown separately. Some of the commands, show the bandwidth used by individual processes. This makes it easy
Linux Network Namespace ã§éŠã‚“ã§ã¿ã‚‹
Linux Network Namespace を使ã†ã¨ 1 ã¤ã® Linux ホストã®ä¸ã«ä»®æƒ³çš„ãªãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ç’°å¢ƒã‚’複数作れる。今回ã¯ã‚µãƒ¼ãƒã€ã‚²ãƒ¼ãƒˆã‚¦ã‚§ã‚¤ã€ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆã® 3 ã¤ã®ãƒãƒ¼ãƒ スペースを作ã£ã¦é€šä¿¡ã•ã›ã¦ã¿ã‚‹ã€‚ã“ã®æ©Ÿèƒ½ã‚’使ã†ã«ã¯æ–°ã—ã‚ã® Linux カーãƒãƒ«ã¨ãƒ„ール群ãŒå¿…è¦ã¿ãŸã„ãªã®ã§ãƒ‡ã‚£ã‚¹ãƒˆãƒªãƒ“ューションã«ã¯ Ubuntu 13.04 を使ã£ãŸã€‚
How to achieve Gigabit speeds with Linux
1 Gbit/s network cards have been available for some time now and 10Gbit/s cards have recently become available. However achieving rates of the order of Gigabits per second is not straightforward. It requires careful tuning of several components in the end systems. This document aims to provide information of how to configure end systems to achieve Gbit/s data transfers. Before you begin with the c
syncookiesã¨Listenãƒãƒƒã‚¯ãƒã‚°ã¨ãƒ‘ケットãƒã‚¹ãƒˆ
ãŸã¾ã«ã¯Linuxãƒã‚¿ã‚’。 Listenãƒãƒƒã‚¯ãƒã‚°ã¯ã€ä¼çµ±çš„ãªUNIXã®å®Ÿè£…ã ã¨ã€SYN_RCVDã¨ESTABLISHEDã®ä¸¡æ–¹ã®ã‚½ã‚±ãƒƒãƒˆæ•°ã‚’æ•°ãˆã¾ã™ãŒã€Linuxã®ãã‚Œã¯ESTABLISHEDãªçŠ¶æ…‹ã®æ•°ã ã‘ã‚’æ•°ãˆã‚‹ã‚ˆã†ã«ãªã£ã¦ã„ã¾ã™(manを見よ)。 ã“ã‚Œã¯ä½•ã§ã‹ã¨ã„ã†ã¨ã€ã„ã‚ゆるSYN Flooding攻撃ã¸ã®å¯¾å¿œã¨ã—ã¦ã€Linuxã¯syncookieを実装ã—ãŸã“ã¨ã®å‰¯ä½œç”¨ãªã®ã ã¨æ€ã„ã¾ã™ã€‚syncookieを実装ã—ã¦ã„ã‚‹ã¨ã€SYNã«å¯¾ã—ã¦SYN_ACK(COOKIE)ã‚’è¿”ã™ã‚³ã‚¹ãƒˆãŒã»ã¼ã‚¼ãƒï¼ˆãƒ¡ãƒ¢ãƒªã‚³ã‚¹ãƒˆã¨ã—ã¦ã¯ï¼‰ã«ãªã‚‹ã®ã§ã€æ¥ãŸSYNã«ã™ã¹ã¦SYN+ACKã‚’è¿”ã™ã“ã¨ãŒå¯èƒ½ã§ã™ã€‚ ã—ãŸãŒã£ã¦ã€SYN_RCVDã®æ•°ã¯æ•°ãˆã¦ã‚‚æ„味ãŒãªããªã£ãŸã®ã§ã€ãã‚Œã¯Listenãƒãƒƒã‚¯ãƒã‚°ã®æ•°ã¨ã—ã¦ã‚«ã‚¦ãƒ³ãƒˆã—ãªã„よã†ã«ã—ãŸã€ã¨ã„ã†ã“ã¨ã®ã‚ˆã†ã§ã™ï¼ˆåŽ³å¯†ã«è¨€ã†ã¨ã€tcp_max_syn_backlog個ã¾ã§
Deprecated Linux networking commands and their replacements
In my article detailing the command line utilities available for configuring and troubleshooting network properties on Windows and Linux, I mentioned some Linux tools that, while still included and functional in many Linux distributions, are actually considered deprecated and therefore should be phased out in favor of more modern replacements. Specifically, the deprecated Linux networking commands
TCPカーãƒãƒ«ãƒ‘ラメータã«ã‚ˆã‚‹éšœå®³å¾©æ—§æ™‚é–“ã®çŸç¸® - GeekFactory
クラスタ構æˆã®ã‚µãƒ¼ãƒã§ã¯ã€éšœå®³ç™ºç”Ÿå¾Œã«ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆãŒã™ãã«å¾©æ—§ã—ãªã„å ´åˆãŒã‚ã‚Šã¾ã™ã€‚サーãƒå´ãŒãƒ•ã‚§ã‚¤ãƒ«ã‚ªãƒ¼ãƒã—ãŸå¾Œã«ã‚¯ãƒ©ã‚¤ã‚¢ãƒ³ãƒˆå´ãŒå†æŽ¥ç¶šã™ã‚‹ã¾ã§ã®æ™‚é–“ã‚’çŸãã™ã‚‹æ–¹æ³•ã‚’紹介ã—ã¾ã™ã€‚ クライアントã‹ã‚‰ã‚µãƒ¼ãƒã«æŽ¥ç¶šã™ã‚‹ã¨ã‚½ã‚±ãƒƒãƒˆã¯ESTABLISHEDã«ãªã‚Šã¾ã™ã€‚ã‚‚ã—ESTABLISHEDã«ãªã£ãŸã‚½ã‚±ãƒƒãƒˆã§æ£ã—ãパケットãŒé€ä¿¡ã•ã‚Œãªã‹ã£ãŸå ´åˆã€OSã¯å†é€ã‚’試ã¿ã¾ã™ã€‚å†é€ã«å¤±æ•—ã—ã¦ã‚½ã‚±ãƒƒãƒˆã‚’クãƒãƒ¼ã‚ºã™ã‚‹ã¾ã§ã®æ™‚é–“ã¯OSã®è¨å®šã«ã‚ˆã‚Šã¾ã™ã€‚ OSãŒTCP接続ã®ç•°å¸¸ã‚’検知ã—ã¦ã‹ã‚‰ã‚¯ãƒãƒ¼ã‚ºã™ã‚‹ã¾ã§ã®æ™‚é–“ã‚’çŸãã™ã‚‹ã«ã¯3ã¤ã®æ–¹æ³•ãŒã‚ã‚Šã¾ã™ã€‚ パケットã®å†é€å›žæ•°ã‚’å°‘ãªãã™ã‚‹ã€‚ TCPレイヤã§Keep Aliveパケットをé€ä¿¡ã™ã‚‹ã€‚ã“ã®æ–¹æ³•ã¯TCP Keep Aliveã«å¯¾å¿œã—ã¦ã„るアプリã®ã¿å¯èƒ½ã€‚ アプリケーションレイヤã§Keep Aliveパケットをé€ä¿¡ã™ã‚‹ã€‚ã“ã®æ–¹æ³•ã¯Nullパケットを投ã’ã‚‹ç‰ã«å¯¾å¿œã—ã¦
TCP Fast Open: expediting web services [LWN.net]
Much of today's Internet traffic takes the form of short TCP data flows that consist of just a few round trips exchanging data segments before the connection is terminated. The prototypical example of this kind of short TCP conversation is the transfer of web pages over the Hypertext Transfer Protocol (HTTP). The speed of TCP data flows is dependent on two factors: transmission delay (the width of
イマドã‚ãªNetwork/IO
The document discusses techniques for improving network and I/O performance between the network interface card (NIC) and CPU. It describes technologies like TCP offloading, receive side scaling, and Intel I/O acceleration features that distribute processing load away from the CPU to improve throughput and reduce latency. Optimization goals include more efficiently handling interrupts and direct me
CentOS 6.2 㧠RPS/RFS を使ã£ã¦ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã®å‰²ã‚Šè¾¼ã¿å‡¦ç†ã‚’複数コアã«åˆ†æ•£ã—ã¦ã¿ãŸ - blog.nomadscafe.jp
以å‰(2010å¹´)ã«ã€Œã‚¢ãƒ—リケーションãŒãƒžãƒ«ãƒã‚¹ãƒ¬ãƒƒãƒ‰ã§ã‚‚マルãƒã‚³ã‚¢CPUã‚’æ´»ã‹ã›ãªã„件ã€ã¨ã„ã†ã‚¨ãƒ³ãƒˆãƒªã«ã¦CPUã®ã‚³ã‚¢ãŒå¢—ãˆã¦ã‚‚割り込ã¿å‡¦ç†ãŒåˆ†æ•£ã•ã‚Œãªã„ã®ã§ã‚¹ã‚±ãƒ¼ãƒ«ã•ã‚Œãªã„ã¨æ›¸ã„ãŸã‘ã©ã€ãã®å¾ŒLinux Kernelã«RPS/RFSãªã‚‹æ©Ÿèƒ½ãŒè¿½åŠ ã•ã‚Œã€å‰²ã‚Šè¾¼ã¿å‡¦ç†ãŒåˆ†æ•£ã§ãるよã†ã«ãªã‚Šã€CentOS 6.2 ã§ã‚‚使ãˆã‚‹ã‚‰ã—ã„ã®ã§è©¦ã—ã¦ã¿ã¾ã—ãŸã€‚ RPS/RFSã«ã¤ã„ã¦ã®ç´¹ä»‹ã¯ VIOPS06ã§ã€ŒRPS・RFSç‰æœ€æ–°Linux Kernel事例ã€ã¨é¡Œã—ã¦ãŠè©±ã—ã¦ãã¾ã—㟠http://d.hatena.ne.jp/syuu1228/20110722/1311322653 Linuxå†…æ ¸ RPS/RFSåŠŸèƒ½è¯¦ç»†æµ‹è¯•åˆ†æž http://www.igigo.net/archives/204 ãŒè©³ã—ã„。2番目ã®ã¯ã»ã¼èªã‚ãªã„ã‘ã©ã€æ€§èƒ½æ¯”較ã®ã‚°ãƒ©ãƒ•ãŒåˆ†ã‹ã‚Šã‚„ã™ã„。 今回試ã—ãŸã‚µãƒ¼ãƒã¯ã€ OS: C
Linuxã®bonding(802.3ad)ã§ç™ºç”Ÿã—ãŸãƒˆãƒ©ãƒ•ã‚£ãƒƒã‚¯ã®åã‚Šã‚’ãªãŠã—ãŸè©± - å…ƒRX-7ä¹—ã‚Šã®é©å½“ãªæ—¥ã€…
ã¯ã˜ã‚ã« ã¨ã‚る環境ã®è©±ã€‚internalã®Linuxサーãƒã§bonding(ボンディング)を組んã§ã„ã¾ã—ãŸã€‚modeã¯4。802.3ad(LACP)æº–æ‹ ã®ãƒªãƒ³ã‚¯ã‚¢ã‚°ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ãªãƒ¢ãƒ¼ãƒ‰ã§ã™ã€‚ ã¡ãªã¿ã«ã€bondingã¨ã¯ãƒ»ãƒ»ãƒ» ã¡ãªã¿ã«ã€"bonding"ã¨ã¯ã€ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã‚¤ãƒ³ã‚¿ãƒ¼ãƒ•ã‚§ãƒ¼ã‚¹ã‚’冗長化(ã¾ãŸã¯è² è·åˆ†æ•£)ã™ã‚‹æ–¹æ³•ã§ã€è¤‡æ•°ã®NICã‚’æŸãã¦1本ã«è¦‹ã›ã‚‹ã“ã¨ãŒã§ãã¾ã™ã€‚ãƒãƒ¼ãƒŸãƒ³ã‚°(teaming)ã¨å‘¼ã°ã‚ŒãŸã‚Šã‚‚ã—ã¾ã™ã。 ã§ã€Linuxã§ã¯bondingã«ã‚‚ã„ãã¤ã‹ãƒ¢ãƒ¼ãƒ‰ãŒã‚ã£ã¦ã€è¤‡æ•°ã®ãƒãƒªã‚·ãƒ¼ã®ä¸ã‹ã‚‰é¸æŠžã™ã‚‹ã“ã¨ãŒã§ãã¾ã™ã€‚ balance-rr åˆã¯ 0 - è€éšœå®³æ€§ã¨ãƒãƒ¼ãƒ‰ãƒãƒ©ãƒ³ã‚·ãƒ³ã‚°ã®ãŸã‚ラウンドãƒãƒ“ンãƒãƒªã‚·ãƒ¼ã‚’è¨å®šã—ã¾ã™ã€‚利用å¯èƒ½ãªç¬¬ 1 ã®ã‚¤ãƒ³ã‚¿ãƒ¼ãƒ•ã‚§ãƒ¼ã‚¹ã‹ã‚‰ãã‚Œãžã‚Œã®ãƒœãƒ³ãƒ‡ã‚£ãƒ³ã‚°ã•ã‚ŒãŸã‚¹ãƒ¬ãƒ¼ãƒ–インターフェースã§é€å—ä¿¡ãŒé †æ¬¡è¡Œã‚ã‚Œã¾ã™ã€‚ active-backup åˆã¯
Linux Tuning
This page contains a quick reference guide for Linux tuning for Data Transfer hosts connected at speeds of 1Gbps or higher. Note that most of the tuning settings described here will actually decrease performance of hosts connected at rates less than 1Gbps, such as most home users. Note that the settings on this page are not attempting to achieve full 10G with a single flow. These settings assume y
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
Recent Linux TCP Updates, and how to tune your 100G host
Redirecting...
mPulse — Real User Performance Monitoring and Real-Time Analytics | Akamai
networking:start [Wiki]
{{#tags}}- {{label}}
{{/tags}}