| <!DOCTYPE html> |
| <!--[if lt IE 7]> <html class="no-js lt-ie9 lt-ie8 lt-ie7"> <![endif]--> |
| <!--[if IE 7]> <html class="no-js lt-ie9 lt-ie8"> <![endif]--> |
| <!--[if IE 8]> <html class="no-js lt-ie9"> <![endif]--> |
| <!--[if gt IE 8]><!--> <html class="no-js"> <!--<![endif]--> |
| <head> |
| <title>SystemML Performance Testing - SystemML 1.2.0</title> |
| <meta charset="utf-8"> |
| <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> |
| |
| <meta name="description" content="Description of SystemML performance testing."> |
| |
| <meta name="viewport" content="width=device-width"> |
| <link rel="stylesheet" href="css/bootstrap.min.css"> |
| <link rel="stylesheet" href="css/main.css"> |
| <link rel="stylesheet" href="css/pygments-default.css"> |
| <link rel="shortcut icon" href="img/favicon.png"> |
| </head> |
| <body> |
| <!--[if lt IE 7]> |
| <p class="chromeframe">You are using an outdated browser. <a href="http://browsehappy.com/">Upgrade your browser today</a> or <a href="http://www.google.com/chromeframe/?redirect=true">install Google Chrome Frame</a> to better experience this site.</p> |
| <![endif]--> |
| |
| <header class="navbar navbar-default navbar-fixed-top" id="topbar"> |
| <div class="container"> |
| <div class="navbar-header"> |
| <div class="navbar-brand brand projectlogo"> |
| <a href="http://systemml.apache.org/"><img class="logo" src="img/systemml-logo.png" alt="Apache SystemML" title="Apache SystemML"/></a> |
| </div> |
| <div class="navbar-brand brand projecttitle"> |
| <a href="http://systemml.apache.org/">Apache SystemML<sup id="trademark">™</sup></a><br/> |
| <span class="version">1.2.0</span> |
| </div> |
| <button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target=".navbar-collapse"> |
| <span class="sr-only">Toggle navigation</span> |
| <span class="icon-bar"></span> |
| <span class="icon-bar"></span> |
| <span class="icon-bar"></span> |
| </button> |
| </div> |
| <nav class="navbar-collapse collapse"> |
| <ul class="nav navbar-nav navbar-right"> |
| <li><a href="index.html">Overview</a></li> |
| <li><a href="https://github.com/apache/systemml">GitHub</a></li> |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">Documentation<b class="caret"></b></a> |
| <ul class="dropdown-menu" role="menu"> |
| <li><b>Running SystemML:</b></li> |
| <li><a href="https://github.com/apache/systemml">SystemML GitHub README</a></li> |
| <li><a href="spark-mlcontext-programming-guide.html">Spark MLContext</a></li> |
| <li><a href="spark-batch-mode.html">Spark Batch Mode</a> |
| <li><a href="hadoop-batch-mode.html">Hadoop Batch Mode</a> |
| <li><a href="standalone-guide.html">Standalone Guide</a></li> |
| <li><a href="jmlc.html">Java Machine Learning Connector (JMLC)</a> |
| <li class="divider"></li> |
| <li><b>Language Guides:</b></li> |
| <li><a href="dml-language-reference.html">DML Language Reference</a></li> |
| <li><a href="beginners-guide-to-dml-and-pydml.html">Beginner's Guide to DML and PyDML</a></li> |
| <li><a href="beginners-guide-python.html">Beginner's Guide for Python Users</a></li> |
| <li><a href="python-reference.html">Reference Guide for Python Users</a></li> |
| <li class="divider"></li> |
| <li><b>ML Algorithms:</b></li> |
| <li><a href="algorithms-reference.html">Algorithms Reference</a></li> |
| <li class="divider"></li> |
| <li><b>Tools:</b></li> |
| <li><a href="debugger-guide.html">Debugger Guide</a></li> |
| <li><a href="developer-tools-systemml.html">IDE Guide</a></li> |
| <li class="divider"></li> |
| <li><b>Other:</b></li> |
| <li><a href="contributing-to-systemml.html">Contributing to SystemML</a></li> |
| <li><a href="engine-dev-guide.html">Engine Developer Guide</a></li> |
| <li><a href="troubleshooting-guide.html">Troubleshooting Guide</a></li> |
| <li><a href="release-process.html">Release Process</a></li> |
| </ul> |
| </li> |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">API Docs<b class="caret"></b></a> |
| <ul class="dropdown-menu" role="menu"> |
| <li><a href="./api/java/index.html">Java</a></li> |
| <li><a href="./api/python/index.html">Python</a></li> |
| </ul> |
| </li> |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">Issues<b class="caret"></b></a> |
| <ul class="dropdown-menu" role="menu"> |
| <li><b>JIRA:</b></li> |
| <li><a href="https://issues.apache.org/jira/browse/SYSTEMML">SystemML JIRA</a></li> |
| |
| </ul> |
| </li> |
| </ul> |
| </nav> |
| </div> |
| </header> |
| |
| <div class="container" id="content"> |
| |
| <h1 class="title">SystemML Performance Testing</h1> |
| |
| |
| <!-- |
| |
| --> |
| |
| <ul id="markdown-toc"> |
| <li><a href="#performance-testing-algorithms-user-manual" id="markdown-toc-performance-testing-algorithms-user-manual">Performance Testing Algorithms User Manual</a> <ul> |
| <li><a href="#architecture" id="markdown-toc-architecture">Architecture</a></li> |
| <li><a href="#adding-new-algorithms" id="markdown-toc-adding-new-algorithms">Adding New Algorithms</a></li> |
| <li><a href="#current-default-settings" id="markdown-toc-current-default-settings">Current Default Settings</a></li> |
| <li><a href="#examples" id="markdown-toc-examples">Examples</a></li> |
| <li><a href="#google-sheets-api" id="markdown-toc-google-sheets-api">Google sheets API</a></li> |
| <li><a href="#result-consolidation-and-plotting" id="markdown-toc-result-consolidation-and-plotting">Result Consolidation and Plotting</a></li> |
| <li><a href="#operational-notes" id="markdown-toc-operational-notes">Operational Notes</a></li> |
| <li><a href="#troubleshooting" id="markdown-toc-troubleshooting">Troubleshooting</a></li> |
| </ul> |
| </li> |
| </ul> |
| |
| <h1 id="performance-testing-algorithms-user-manual">Performance Testing Algorithms User Manual</h1> |
| |
| <p>This user manual contains details on how to conduct automated performance tests. Work was mostly done in this <a href="https://github.com/apache/systemml/pull/537">PR</a> and part of <a href="https://issues.apache.org/jira/browse/SYSTEMML-1451">SYSTEMML-1451</a>. Our aim was to move from existing <code>bash</code> based performance tests to automatic <code>python</code> based automatic performance tests.</p> |
| |
| <h2 id="architecture">Architecture</h2> |
| |
| <p>Our performance tests suit contains <code>7</code> families namely <code>binomial</code>, <code>multinomial</code>, <code>stats1</code>, <code>stats2</code>, <code>regression1</code>, <code>regression2</code>, <code>clustering</code>. Within these families we have algorithms grouped under it. Typically a family is a set of algorithms that require the same data generation script.</p> |
| |
| <ul> |
| <li>Exceptions: <code>regression1</code>, <code>regression2</code> and <code>binomial</code>. We decide to include these algorithms in separate families to keep the architecture simple.</li> |
| </ul> |
| |
| <p><img src="img/performance-test/perf_test_arch.png" alt="System ML Architecture" /></p> |
| |
| <p>On a very high level use construct a string with arguments required to run each operation. Once this string is constructed we use the subprocess module to execute this string and extract time from the standard out.</p> |
| |
| <p>We also use <code>json</code> module write our configurations to a json file. This ensure that our operations are easy to debug.</p> |
| |
| <p>We have <code>7</code> files in performance test suit:</p> |
| |
| <ul> |
| <li>Entry File <code>run_perftest.py</code></li> |
| <li>Supporting Files <code>datagen.py</code>, <code>train.py</code>, <code>predict.py</code></li> |
| <li>Utility Files <code>utils_exec.py</code>, <code>utils_fs.py</code>, <code>utils_misc.py</code></li> |
| </ul> |
| |
| <p><code>datagen.py</code>, <code>train.py</code> and <code>predict.py</code> generate a dictionary. Our key is the name of algorithm being processed and values is a list with path(s) where all the data required is present. We define this dictionary as a configuration packet.</p> |
| |
| <p>We will describe each of them in detail the following sections below.</p> |
| |
| <p><code>run_perftest.py</code> at a high level creates <code>algos_to_run</code> list. This list is tuple with key as algorithm and value as family to be executed in our performance test.</p> |
| |
| <p>In <code>datagen.py</code> script we have all functions required to generate data. We return the required configuration packet as a result of this script, that contains key as the <code>data-gen</code> script to run and values with location to read data-gen json files from.</p> |
| |
| <p>In <code>train.py</code> script we have functions required to generate training output. We return the required configuration packet as a result of this script, that contains key as the algorithm to run and values with location to read training json files from.</p> |
| |
| <p>The file <code>predict.py</code> contains all functions for all algorithms in the performance test that contain predict script. We return the required configuration packet as a result of this script, that contains key as the algorithm to run and values with location to read predict json files from.</p> |
| |
| <p>In the file(s) <code>utils_*.py</code> we have all the helper functions required in our performance test. These functions do operations like write <code>json</code> files, extract time from std out etc.</p> |
| |
| <h2 id="adding-new-algorithms">Adding New Algorithms</h2> |
| |
| <p>While adding a new algorithm we need know if it has to be part of the any pre existing family. If this algorithm depends on a new data generation script we would need to create a new family. Steps below to take below to add a new algorithm.</p> |
| |
| <p>Following changes to <code>run_perftest.py</code>:</p> |
| |
| <ul> |
| <li>Add the algorithm to <code>ML_ALGO</code> dictionary with its respective family.</li> |
| <li>Add the name of the data generation script in <code>ML_GENDATA</code> dictionary if it does not exist already.</li> |
| <li>Add the name of the training script in <code>ML_TRAIN</code> dictionary.</li> |
| <li>Add the name of the prediction script in <code>ML_PREDICT</code> incase the prediction script exists.</li> |
| </ul> |
| |
| <p>Following changes to <code>datagen.py</code>:</p> |
| |
| <ul> |
| <li>Check if the data generation algorithm has the ability to generate dense and sparse data. If it had the ability to generate only dense data add the corresponding family to <code>FAMILY_NO_MATRIX_TYPE</code> list.</li> |
| <li>Create a function with <code>familyname + _ + datagen</code> with same input arguments namely <code>matrix_dim</code>, <code>matrix_type</code>, <code>datagen_dir</code>.</li> |
| <li>Constants and arguments for the data generation script should be defined in function.</li> |
| <li>Test the perf test with the algorithm with <code>mode</code> as <code>data-gen</code>.</li> |
| <li>Check output folders, json files, output log.</li> |
| <li>Check for possible errors if these folders/files do not exist. (See the troubleshooting section).</li> |
| </ul> |
| |
| <p>Following changes to <code>train.py</code>:</p> |
| |
| <ul> |
| <li>Create the function with <code>familyname + _ + algoname + _ + train</code>.</li> |
| <li>This function needs to have the following arguments <code>save_folder_name</code>, <code>datagen_dir</code>, <code>train_dir</code>.</li> |
| <li>Constants and arguments for the training script should be defined in function.</li> |
| <li>Make sure that the return type is a list.</li> |
| <li>Test the perf test with the algorithm with <code>mode</code> as <code>train</code>.</li> |
| <li>Check output folders, json files, output log.</li> |
| <li>Check for possible errors if these folders/files do not exist. (See the troubleshooting section).</li> |
| </ul> |
| |
| <p>Following changes to <code>predict.py</code>:</p> |
| |
| <ul> |
| <li>Create the function with <code>algoname + _ + predict</code>.</li> |
| <li>This function needs to have the following arguments <code>save_file_name</code>, <code>datagen_dir</code>, <code>train_dir</code>, <code>predict_dir</code>.</li> |
| <li>Constants and arguments for the training script should be defined in function.</li> |
| <li>Test the perf test with the algorithm with <code>mode</code> as <code>predict</code>.</li> |
| <li>Check output folders, json files, output log.</li> |
| <li>Check for possible errors if these folders/files do not exist. (Please see the troubleshooting section).</li> |
| <li>Note: <code>predict.py</code> will not be executed if the current algorithm being executed does not have predict script.</li> |
| </ul> |
| |
| <h2 id="current-default-settings">Current Default Settings</h2> |
| |
| <p>Default setting for our performance test below:</p> |
| |
| <ul> |
| <li>Matrix size to 10,000 rows and 100 columns.</li> |
| <li>Execution mode <code>singlenode</code>.</li> |
| <li>Operation modes <code>data-gen</code>, <code>train</code> and <code>predict</code> in sequence.</li> |
| <li>Matrix type set to <code>all</code>. Which will generate <code>dense</code>, <code>sparse</code> matrices for all relevant algorithms.</li> |
| </ul> |
| |
| <h2 id="examples">Examples</h2> |
| |
| <p>Some examples of SystemML performance test with arguments shown below:</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --family binomial clustering multinomial regression1 regression2 stats1 stats2 |
| </code> |
| Test all algorithms with default parameters.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --exec-type hybrid_spark --family binomial clustering multinomial regression1 regression2 stats1 stats2 |
| </code> |
| Test all algorithms in hybrid spark execution mode.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --exec-type hybrid_spark --family clustering --mat-shape 10k_5 10k_10 10k_50 |
| </code> |
| Test all algorithms in <code>clustering</code> family in hybrid spark execution mode, on different matrix size <code>10k_10</code> (10,000 rows and 5 columns), <code>10k_10</code> and <code>10k_50</code>.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --algo Univar-Stats bivar-stats |
| </code> |
| Run performance test for following algorithms <code>Univar-Stats</code> and <code>bivar-stats</code>.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --algo m-svm --family multinomial binomial --mode data-gen train |
| </code> |
| Run performance test for the algorithms <code>m-svm</code> with <code>multinomial</code> family. Run only data generation and training operations.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --family regression2 --filename new_log |
| </code> |
| Run performance test for all algorithms under the family <code>regression2</code> and log with filename <code>new_log</code>.</p> |
| |
| <p><code>./scripts/perftest/python/run_perftest.py --family binomial clustering multinomial regression1 regression2 stats1 stats2 --config-dir /Users/krishna/open-source/systemml/scripts/perftest/temp3 --temp-dir hdfs://localhost:9000/temp3</code> |
| Run performance test for all algorithms using HDFS.</p> |
| |
| <h2 id="google-sheets-api">Google sheets API</h2> |
| |
| <p>Steps below to configure google client API:</p> |
| |
| <ul> |
| <li>Navigate to <a href="https://console.developers.google.com/apis/">Google APIs Console</a>.</li> |
| <li>Create a new project.</li> |
| <li>Click Enable API. Search for and enable the Google Drive API.</li> |
| <li>Create credentials for a Web Server to access Application Data.</li> |
| <li>Name the service account and grant it a Project Role of Editor.</li> |
| <li>Download the JSON file.</li> |
| <li>Copy the JSON file to your code directory and rename it to client_secret.json</li> |
| </ul> |
| |
| <p>Steps below to configure google sheets:</p> |
| |
| <ul> |
| <li>Create a new spread sheet with google sheets.</li> |
| <li>Create seperate sheets for <code>singlenode</code> and <code>hybrid_spark</code>.</li> |
| <li>Find the client_email inside client_secret.json and save it.</li> |
| <li>Back in your spreadsheet, click the Share button in the top right, and paste the client email into the People field to give it edit rights for each sheet.</li> |
| <li>Click Send</li> |
| </ul> |
| |
| <h2 id="result-consolidation-and-plotting">Result Consolidation and Plotting</h2> |
| <p>We have two scripts, <code>stats.py</code> forpulling results from google docs and <code>update.py</code> to updating results to google docs or local file system.</p> |
| |
| <p>Example of <code>update.py</code> would be below |
| <code>./scripts/perftest/python/google_docs/update.py --file ../../temp/perf_test_singlenode.out --exec-type singlenode --tag 2 --append test.csv</code> |
| The arguments being <code>--file</code> path of the perf-test output, <code>--exec-type</code> execution mode used to generate the perf-test output, <code>--tag</code> being the realease version or a unique name, <code>--append</code> being an optional argument that would append the a local csv file. If instead of <code>--append</code> the <code>--auth</code> argument needs the location of the <code>google api key</code> file.</p> |
| |
| <p>Example of <code>stats.py</code> below |
| ` ./stats.py –auth ../key/client_json.json –exec-type singlenode –plot stats1_data-gen_none_dense_10k_100<code> |
| </code>–plot` argument needs the name of the composite key that you would like to compare results over. If this argument is not specified the results would be grouped by keys.</p> |
| |
| <h2 id="operational-notes">Operational Notes</h2> |
| |
| <p>All performance test depend mainly on two scripts for execution <code>systemml-standalone.py</code> and <code>systemml-spark-submit.py</code>. Incase we need to change standalone or spark parameters we need to manually change these parameters in their respective scripts.</p> |
| |
| <p>Constants like <code>DATA_FORMAT</code> currently set to <code>csv</code> and <code>MATRIX_TYPE_DICT</code> with <code>density</code> set to <code>0.9</code> and <code>sparsity</code> set to <code>0.01</code> are hardcoded in the performance test scripts. They can be changed easily as they are defined at the top of their respective operational scripts.</p> |
| |
| <p>The logs contain the following information below comma separated.</p> |
| |
| <table> |
| <thead> |
| <tr> |
| <th>algorithm</th> |
| <th>run_type</th> |
| <th>intercept</th> |
| <th>matrix_type</th> |
| <th>data_shape</th> |
| <th>time_sec</th> |
| </tr> |
| </thead> |
| <tbody> |
| <tr> |
| <td>multinomial</td> |
| <td>data-gen</td> |
| <td>0</td> |
| <td>10k_100</td> |
| <td>dense</td> |
| <td>0.33</td> |
| </tr> |
| <tr> |
| <td>MultiLogReg</td> |
| <td>train</td> |
| <td>0</td> |
| <td>10k_100</td> |
| <td>dense</td> |
| <td>6.956</td> |
| </tr> |
| <tr> |
| <td>MultiLogReg</td> |
| <td>predict</td> |
| <td>0</td> |
| <td>10k_100</td> |
| <td>dense</td> |
| <td>4.780</td> |
| </tr> |
| </tbody> |
| </table> |
| |
| <p>These logs and config <code>json</code> files can be found in <code>temp</code> folder (<code>$SYSTEMML_HOME/scripts/perftest/temp</code>) in-case not overridden by <code>--config-dir</code>.</p> |
| |
| <p><code>--temp-dir</code> by default points to local file system. We can change this to point to a hdfs path by <code>--temp-dir hdfs://localhost:9000/temp</code> where all files generated during execution will be saved.</p> |
| |
| <p>Every time a script executes in <code>data-gen</code> mode successfully, we write a <code>_SUCCESS</code> file. If this file exists we ensures that re-runs of the same script is not possible. Support for configuration options like <code>-stats</code>, <code>-explain</code>, <code>--conf</code> have also been added.</p> |
| |
| <p>Results obtained by our performance tests can be automatically uploaded to google docs.</p> |
| |
| <p><code>./update.py --file ../temp/singlenode.out --exec-mode singlenode --auth client_json.json --tag 1.0</code></p> |
| |
| <p>In the example above <code>--tag</code> can be a major/minor systemml version and <code>--auth</code> points to the <code>json</code> key required by <code>google docs</code>.</p> |
| |
| <p>Currently we only support time difference between algorithms in different versions. This can be obtained by running the script below |
| <code>./stats.py --auth client_json.json --exec-mode singlenode --tags 1.0 2.0</code></p> |
| |
| <p>We pass different <code>matrix shapes</code> using <code>--mat-shape</code> argument.</p> |
| |
| <table> |
| <thead> |
| <tr> |
| <th>Matrix Shape</th> |
| <th>Approximate Data Size</th> |
| </tr> |
| </thead> |
| <tbody> |
| <tr> |
| <td>10k_1k</td> |
| <td>80MB</td> |
| </tr> |
| <tr> |
| <td>100k_1k</td> |
| <td>800MB</td> |
| </tr> |
| <tr> |
| <td>1M_1k</td> |
| <td>8GB</td> |
| </tr> |
| <tr> |
| <td>10M_1k</td> |
| <td>80GB</td> |
| </tr> |
| <tr> |
| <td>100M_1k</td> |
| <td>800GB</td> |
| </tr> |
| </tbody> |
| </table> |
| |
| <p>For example the command below runs performance test for all data sizes described above |
| <code>run_perftest.py --family binomial clustering multinomial regression1 regression2 stats1 stats2 --mat-shape 10k_1k 100k_1k 1M_1k 10M_1k 100M_1k --master yarn-client --temp-dir hdfs://localhost:9000/user/systemml</code></p> |
| |
| <p>By default data generated in <code>hybrid_spark</code> execution mode is in the current users <code>hdfs</code> home directory.</p> |
| |
| <p>Note: Please use this command <code>pip3 install -r requirements.txt</code> before using the perftest scripts.</p> |
| |
| <h2 id="troubleshooting">Troubleshooting</h2> |
| |
| <p>We can debug the performance test by making changes in the following locations based on</p> |
| |
| <ul> |
| <li>Please see <code>utils_exec.py</code> function <code>subprocess_exec</code>.</li> |
| <li>Please see <code>run_perftest.py</code>. Changing the verbosity level to <code>0</code> allows us to log more information while the script runs.</li> |
| <li>Eyeballing the json files generated and making sure the configuration arguments are correct.</li> |
| </ul> |
| |
| |
| </div> <!-- /container --> |
| |
| |
| |
| <script src="js/vendor/jquery-1.12.0.min.js"></script> |
| <script src="js/vendor/bootstrap.min.js"></script> |
| <script src="js/vendor/anchor.min.js"></script> |
| <script src="js/main.js"></script> |
| |
| |
| |
| |
| |
| <!-- Analytics --> |
| <script> |
| (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ |
| (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), |
| m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) |
| })(window,document,'script','//www.google-analytics.com/analytics.js','ga'); |
| ga('create', 'UA-71553733-1', 'auto'); |
| ga('send', 'pageview'); |
| </script> |
| |
| |
| |
| <!-- MathJax Section --> |
| <script type="text/x-mathjax-config"> |
| MathJax.Hub.Config({ |
| TeX: { equationNumbers: { autoNumber: "AMS" } } |
| }); |
| </script> |
| <script> |
| // Note that we load MathJax this way to work with local file (file://), HTTP and HTTPS. |
| // We could use "//cdn.mathjax...", but that won't support "file://". |
| (function(d, script) { |
| script = d.createElement('script'); |
| script.type = 'text/javascript'; |
| script.async = true; |
| script.onload = function(){ |
| MathJax.Hub.Config({ |
| tex2jax: { |
| inlineMath: [ ["$", "$"], ["\\\\(","\\\\)"] ], |
| displayMath: [ ["$$","$$"], ["\\[", "\\]"] ], |
| processEscapes: true, |
| skipTags: ['script', 'noscript', 'style', 'textarea', 'pre'] |
| } |
| }); |
| }; |
| script.src = ('https:' == document.location.protocol ? 'https://' : 'http://') + |
| 'cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML'; |
| d.getElementsByTagName('head')[0].appendChild(script); |
| }(document)); |
| </script> |
| </body> |
| </html> |
