[{"_1":2,"_168":-5,"_169":-5},"loaderData",{"_3":4,"_13":14,"_21":22,"_144":145},"root",{"_5":-5,"_6":-7,"_7":8,"_9":-7,"_10":11,"_12":-7},"currentUser","primaryEmail","siteVerification","6LeIxAcTAAAAAJcZVRqyHh71UMIEGNQ_MXjiZKhI","toast","enableDev",false,"incidentStatus","routes/_layout",{"_15":-5,"_16":17,"_18":17,"_19":11,"_20":11},"username","isAskGordonEnabled",true,"isUsingNewTemplate","showMinimalLanding","showHoneypotTrap","routes/_layout.r.$namespace.$reponame",{"_23":24,"_25":-5,"_26":11,"_27":11,"_28":29,"_30":31,"_32":-5,"_33":11,"_34":35,"_85":-5,"_86":87,"_137":-5,"_138":11,"_139":17,"_140":11,"_141":-5,"_142":143},"canonicalUrl","https://hub.docker.com/r/bplotka/mutilate/tags","mcpServer","canSeeDockerfile","canSeeAutobuilds","namespace","bplotka","reponame","mutilate","cloudBuildSource","isAutobuildsAvailable","repo",{"_36":37,"_38":39,"_41":42,"_43":44,"_45":11,"_46":47,"_48":49,"_50":51,"_53":31,"_28":29,"_54":55,"_56":-5,"_57":58,"_59":60,"_61":62,"_63":-5,"_64":65,"_66":67,"_68":11,"_69":29,"_70":71,"_76":17,"_77":78,"_82":29,"_83":84},"categories",[],"content_types",[40],"image","date_registered","2016-01-14T17:58:13.8758Z","description","Improved mutilate in docker.","is_private","last_updated","2016-02-03T15:35:23.948109Z","last_modified","2024-10-16T13:48:34.145251Z","media_types",[52],"application/vnd.docker.distribution.manifest.v1+prettyjws","name","pull_count",63123,"repository_type","star_count",2,"status",1,"status_description","active","source","collaborator_count",0,"full_description","Improved Mutilate\n========\n \nMutilate is a memcached load generator designed for high request\nrates, good tail-latency measurements, and realistic request stream\ngeneration.\n\nImprovements\n============\n\n1. Masterless mode.\n2. Sinusoidal type of stress [WIP]\n3. Storing latency of each request to InfluxDB [WIP]\n\nRequirements\n============\n\n1. A C++0x compiler\n2. scons\n3. libevent\n4. gengetopt\n5. zeromq (optional)\n\nMutilate has only been thoroughly tested on Ubuntu 11.10. We'll flesh\nout compatibility over time.\n\nBuilding\n========\n\n apt-get install scons libevent-dev gengetopt libzmq-dev\n scons\n\nBasic Usage\n===========\n\nType './mutilate -h' for a full list of command-line options. At\nminimum, a server must be specified.\n\n $ ./mutilate -s localhost\n #type avg min 1st 5th 10th 90th 95th 99th\n read 52.4 41.0 43.1 45.2 48.1 55.8 56.6 71.5\n update 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0\n op_q 1.5 1.0 1.0 1.1 1.1 1.9 2.0 2.0\n \n Total QPS = 18416.6 (92083 / 5.0s)\n \n Misses = 0 (0.0%)\n \n RX 22744501 bytes : 4.3 MB/s\n TX 3315024 bytes : 0.6 MB/s\n\nMutilate reports the latency (average, minimum, and various\npercentiles) for get and set commands, as well as achieved QPS and\nnetwork goodput.\n\nTo achieve high request rate, you must configure mutilate to use\nmultiple threads, multiple connections, connection pipelining, or\nremote agents.\n\n $ ./mutilate -s zephyr2-10g -T 24 -c 8\n #type avg min 1st 5th 10th 90th 95th 99th\n read 598.8 86.0 437.2 466.6 482.6 977.0 1075.8 1170.6\n update 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0\n op_q 1.5 1.0 1.0 1.1 1.1 1.9 1.9 2.0\n \n Total QPS = 318710.8 (1593559 / 5.0s)\n \n Misses = 0 (0.0%)\n \n RX 393609073 bytes : 75.1 MB/s\n TX 57374136 bytes : 10.9 MB/s\n\nSuggested Usage\n===============\n\nReal deployments of memcached often handle the requests of dozens,\nhundreds, or thousands of front-end clients simultaneously. However,\nby default, mutilate establishes one connection per server and meters\nrequests one at a time (it waits for a reply before sending the next\nrequest). This artificially limits throughput (i.e. queries per\nsecond), as the round-trip network latency is almost certainly far\nlonger than the time it takes for the memcached server to process one\nrequest.\n\nIn order to get reasonable benchmark results with mutilate, it needs\nto be configured to more accurately portray a realistic client\nworkload. In general, this means ensuring that (1) there are a large\nnumber of client connections, (2) there is the potential for a large\nnumber of outstanding requests, and (3) the memcached server saturates\nand experiences queuing delay far before mutilate does. I suggest the\nfollowing guidelines:\n\n1. Establish on the order of 100 connections per memcached _server_\nthread.\n2. Don't exceed more than about 16 connections per mutilate thread.\n3. Use multiple mutilate agents in order to achieve (1) and (2).\n4. Do not use more mutilate threads than hardware cores/threads.\n5. Use -Q to configure the \"master\" agent to take latency samples at\nslow, a constant rate.\n\nHere's an example:\n\n memcached_server$ memcached -t 4 -c 32768\n agent1$ mutilate -T 16 -A\n agent2$ mutilate -T 16 -A\n agent3$ mutilate -T 16 -A\n agent4$ mutilate -T 16 -A\n agent5$ mutilate -T 16 -A\n agent6$ mutilate -T 16 -A\n agent7$ mutilate -T 16 -A\n agent8$ mutilate -T 16 -A\n master$ mutilate -s memcached_server --loadonly\n master$ mutilate -s memcached_server --noload \\\n -B -T 16 -Q 1000 -D 4 -C 4 \\\n -a agent1 -a agent2 -a agent3 -a agent4 \\\n -a agent5 -a agent6 -a agent7 -a agent8 \\\n -c 4 -q 200000\n\nThis will create 8*16*4 = 512 connections total, which is about 128\nper memcached server thread. This ought to be enough outstanding\nrequests to cause server-side queuing delay, and no possibility of\nclient-side queuing delay adulterating the latency measurements.\n\nCommand-line Options\n====================\n\n mutilate3 0.1\n \n Usage: mutilate -s server[:port] [options]\n \n \"High-performance\" memcached benchmarking tool\n \n -h, --help Print help and exit\n --version Print version and exit\n -v, --verbose Verbosity. Repeat for more verbose.\n --quiet Disable log messages.\n \n Basic options:\n -s, --server=STRING Memcached server hostname[:port]. Repeat to \n specify multiple servers.\n --binary Use binary memcached protocol instead of ASCII.\n -q, --qps=INT Target aggregate QPS. 0 = peak QPS. \n (default=`0')\n -t, --time=INT Maximum time to run (seconds). (default=`5')\n -K, --keysize=STRING Length of memcached keys (distribution). \n (default=`30')\n -V, --valuesize=STRING Length of memcached values (distribution). \n (default=`200')\n -r, --records=INT Number of memcached records to use. If \n multiple memcached servers are given, this \n number is divided by the number of servers. \n (default=`10000')\n -u, --update=FLOAT Ratio of set:get commands. (default=`0.0')\n \n Advanced options:\n -U, --username=STRING Username to use for SASL authentication.\n -P, --password=STRING Password to use for SASL authentication.\n -T, --threads=INT Number of threads to spawn. (default=`1')\n --affinity Set CPU affinity for threads, round-robin\n -c, --connections=INT Connections to establish per server. \n (default=`1')\n -d, --depth=INT Maximum depth to pipeline requests. \n (default=`1')\n -R, --roundrobin Assign threads to servers in round-robin \n fashion. By default, each thread connects to \n every server.\n -i, --iadist=STRING Inter-arrival distribution (distribution). \n Note: The distribution will automatically be \n adjusted to match the QPS given by --qps. \n (default=`exponential')\n -S, --skip Skip transmissions if previous requests are \n late. This harms the long-term QPS average, \n but reduces spikes in QPS after long latency \n requests.\n --moderate Enforce a minimum delay of ~1/lambda between \n requests.\n --noload Skip database loading.\n --loadonly Load database and then exit.\n -B, --blocking Use blocking epoll(). May increase latency.\n --no_nodelay Don't use TCP_NODELAY.\n -w, --warmup=INT Warmup time before starting measurement.\n -W, --wait=INT Time to wait after startup to start \n measurement.\n --save=STRING Record latency samples to given file.\n --search=N:X Search for the QPS where N-order statistic < \n Xus. (i.e. --search 95:1000 means find the \n QPS where 95% of requests are faster than \n 1000us).\n --scan=min:max:step Scan latency across QPS rates from min to max.\n \n Agent-mode options:\n -A, --agentmode Run client in agent mode.\n -M, --masterless_agentmode Run client in masterless agent mode.\n -a, --agent=host Enlist remote agent.\n -p, --agent_port=STRING Agent port. (default=`5556')\n -l, --lambda_mul=INT Lambda multiplier. Increases share of QPS for \n this client. (default=`1')\n -C, --measure_connections=INT Master client connections per server, overrides \n --connections.\n -Q, --measure_qps=INT Explicitly set master client QPS, spread across \n threads and connections.\n -D, --measure_depth=INT Set master client connection depth.\n \n The --measure_* options aid in taking latency measurements of the\n memcached server without incurring significant client-side queuing\n delay. --measure_connections allows the master to override the\n --connections option. --measure_depth allows the master to operate as\n an \"open-loop\" client while other agents continue as a regular\n closed-loop clients. --measure_qps lets you modulate the QPS the\n master queries at independent of other clients. This theoretically\n normalizes the baseline queuing delay you expect to see across a wide\n range of --qps values.\n \n Some options take a 'distribution' as an argument.\n Distributions are specified by [:[,...]].\n Parameters are not required. The following distributions are supported:\n \n [fixed:] Always generates .\n uniform: Uniform distribution between 0 and .\n normal:, Normal distribution.\n exponential: Exponential distribution.\n pareto:,, Generalized Pareto distribution.\n gev:,, Generalized Extreme Value distribution.\n \n To recreate the Facebook \"ETC\" request stream from [1], the\n following hard-coded distributions are also provided:\n \n fb_value = a hard-coded discrete and GPareto PDF of value sizes\n fb_key = \"gev:30.7984,8.20449,0.078688\", key-size distribution\n fb_ia = \"pareto:0.0,16.0292,0.154971\", inter-arrival time dist.\n \n [1] Berk Atikoglu et al., Workload Analysis of a Large-Scale Key-Value Store,\n SIGMETRICS 2012\n \n","has_starred","hub_user","immutable_tags_settings",{"_72":11,"_73":74},"enabled","rules",[75],".*","is_automated","permissions",{"_79":17,"_80":11,"_81":11},"read","write","admin","user","storage_size",701778627,"orgData","recentTags",[88,126],{"_89":40,"_90":91,"_92":93,"_94":95,"_96":97,"_46":115,"_116":91,"_117":29,"_118":52,"_53":119,"_120":121,"_122":106,"_123":115,"_124":62,"_125":17},"content_type","creator",479067,"full_size",160798385,"id",1761561,"images",[98],{"_99":100,"_101":102,"_103":104,"_105":106,"_107":108,"_109":104,"_110":104,"_111":-5,"_112":113,"_59":62,"_114":-5},"architecture","amd64","digest","sha256:f3f395b39ae4f543784f8e65286fea8a5cad661b87ad79232944d6974b3e9787","features","","last_pulled","2026-09-16T18:33:09.763514392Z","last_pushed","2016-02-03T15:35:23Z","os","os_features","os_version","size",160798417,"variant","2016-02-03T15:35:22.8463Z","last_updater","last_updater_username","media_type","latest","repository",482012,"tag_last_pulled","tag_last_pushed","tag_status","v2",{"_89":40,"_90":91,"_92":127,"_94":128,"_96":129,"_46":135,"_116":91,"_117":29,"_118":52,"_53":136,"_120":121,"_122":132,"_123":135,"_124":62,"_125":17},160803875,1747169,[130],{"_99":100,"_101":131,"_103":104,"_105":132,"_107":133,"_109":104,"_110":104,"_111":-5,"_112":134,"_59":62,"_114":-5},"sha256:dccc570b2072ec37f388d453ba2cd3ec2305f56f5674b29b0d669ff13006a023","2026-09-17T01:55:40.794231423Z","2016-01-18T16:05:18Z",160803907,"2016-01-18T16:05:17.523465Z","masterless","sbxKitMetadata","isPotentialSpam","isHubSbxKitsEnabled","displaySecurityMetadata","tagAttestations","tagsHaveAttestations",[],"routes/_layout.r.$namespace.$reponame.tags",{"_34":146,"_153":154,"_142":166,"_167":11},{"_36":147,"_38":148,"_41":42,"_43":44,"_45":11,"_46":47,"_48":49,"_50":149,"_53":31,"_28":29,"_54":55,"_56":-5,"_57":58,"_59":60,"_61":62,"_63":-5,"_64":65,"_66":67,"_68":11,"_69":29,"_70":150,"_76":17,"_77":152,"_82":29,"_83":84},[],[40],[52],{"_72":11,"_73":151},[75],{"_79":17,"_80":11,"_81":11},"tags",{"_155":58,"_156":-5,"_157":-5,"_158":159},"count","next","previous","results",[160,163],{"_89":40,"_90":91,"_92":93,"_94":95,"_96":161,"_46":115,"_116":91,"_117":29,"_118":52,"_53":119,"_120":121,"_122":106,"_123":115,"_124":62,"_125":17},[162],{"_99":100,"_101":102,"_103":104,"_105":106,"_107":108,"_109":104,"_110":104,"_111":-5,"_112":113,"_59":62,"_114":-5},{"_89":40,"_90":91,"_92":127,"_94":128,"_96":164,"_46":135,"_116":91,"_117":29,"_118":52,"_53":136,"_120":121,"_122":132,"_123":135,"_124":62,"_125":17},[165],{"_99":100,"_101":131,"_103":104,"_105":132,"_107":133,"_109":104,"_110":104,"_111":-5,"_112":134,"_59":62,"_114":-5},[],"canDeleteTags","actionData","errors"]

bplotka/mutilate

Tags for bplotka/mutilate