{"id":6167,"date":"2016-07-12T18:27:59","date_gmt":"2016-07-12T17:27:59","guid":{"rendered":"http:\/\/www.walkingrandomly.com\/?p=6167"},"modified":"2016-07-12T18:28:54","modified_gmt":"2016-07-12T17:28:54","slug":"high-performance-computing-think-about-where-you-do-data-operations","status":"publish","type":"post","link":"https:\/\/walkingrandomly.com\/?p=6167","title":{"rendered":"High Performance Computing: Think about where you do data operations"},"content":{"rendered":"<p>The <a href=\"http:\/\/docs.iceberg.shef.ac.uk\/en\/latest\/index.html\">High Performance Computing system at University of Sheffield<\/a> has several different file systems available to it. We have:-<\/p>\n<ul>\n<li>\/fastdata &#8211; A <a href=\"https:\/\/www.walkingrandomly.com\/?p=6042\">lustre<\/a>-based, shared filesystem with hundreds of terabytes of space. No backup. No quota.<\/li>\n<li>\/data &#8211; An NFS file system where each user has access to 100Gb of storage. Back-ups go back 7 days.<\/li>\n<li>\/home &#8211; \u00a0An NFS file system where each user has 10Gb. Backed up over 28 days. Mirrored.<\/li>\n<li>\/scratch &#8211; Local disk on each worker node. No back up. Uses ext4.<\/li>\n<\/ul>\n<p>Lots of options with differing amounts of space, back-up policy and, as I&#8217;m about to demonstrate, performance characteristics. I suspect that many other HPC systems have a similar set up.<\/p>\n<p>On our system, it&#8217;s very tempting to do <strong>everything<\/strong>\u00a0in \/fastdata. There&#8217;s lots of space, no quota, readable from all worker nodes simultaneously &#8212; good times! I try to encourage people to think about what they are doing, however. Bad things can happen if the lustre filesystem is hammered too much. Also, there can be a huge difference in performance for some operations across different filesystems.<\/p>\n<p>Let&#8217;s take an example. I want to download and untar gcc 4.9.2. How long does that take on the three different filesystems?<\/p>\n<p><strong>On the scratch directory of a worker node<\/strong><\/p>\n<pre><code>cd \\scratch\r\nmkdir testing123\r\ncd testing123\r\nwget ftp:\/\/ftp.mirrorservice.org\/sites\/sourceware.org\/pub\/gcc\/releases\/gcc-4.9.2\/gcc-4.9.2.tar.gz\r\ntime tar xfz .\/gcc-4.9.2.tar.gz \r\n\r\nreal    0m6.237s\r\nuser    0m5.302s\r\nsys 0m3.033s\r\n<\/code><\/pre>\n<p><strong>On the lustre filesystem<\/strong><\/p>\n<pre><code>cd \/fastdata\/\r\nmkdir testing123\r\ncd testing123\r\nwget ftp:\/\/ftp.mirrorservice.org\/sites\/sourceware.org\/pub\/gcc\/releases\/gcc-4.9.2\/gcc-4.9.2.tar.gz\r\ntime tar xfz .\/gcc-4.9.2.tar.gz \r\n\r\nreal    7m18.170s\r\nuser    0m6.751s\r\nsys 0m56.802s\r\n\r\n<\/code><\/pre>\n<p><strong>On the NFS filesystem<br \/>\n<\/strong><\/p>\n<pre><code>cd \/data\/myusername\r\nmkdir testing123\r\ncd testing123\r\nwget ftp:\/\/ftp.mirrorservice.org\/sites\/sourceware.org\/pub\/gcc\/releases\/gcc-4.9.2\/gcc-4.9.2.tar.gz\r\ntime tar xfz .\/gcc-4.9.2.tar.gz \r\n\r\nreal\t16m37.343s\r\nuser\t0m6.052s\r\nsys\t0m23.438s\r\n<\/code><\/pre>\n<p>For this particular operation, there is a\u00a0<strong>two orders of magnitude<\/strong>\u00a0difference between the worst and the best option.<\/p>\n<p>I&#8217;m not an expert in filesystems and I have no idea what&#8217;s causing these differences or if I&#8217;d see a similar speed difference given a different file operation. I currently have no interest in doing a robust set of benchmarks.\u00a0The point I&#8217;m making is that if you are using a system that has multiple filesystems it may be worth checking if there&#8217;s an advantage to using one over the other for your particular use case.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>The High Performance Computing system at University of Sheffield has several different file systems available to it. We have:- \/fastdata &#8211; A lustre-based, shared filesystem with hundreds of terabytes of space. No backup. No quota. \/data &#8211; An NFS file system where each user has access to 100Gb of storage. Back-ups go back 7 days. [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[68],"tags":[],"class_list":["post-6167","post","type-post","status-publish","format-standard","hentry","category-hpc"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p3swhs-1Bt","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/posts\/6167","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=6167"}],"version-history":[{"count":3,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/posts\/6167\/revisions"}],"predecessor-version":[{"id":6170,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=\/wp\/v2\/posts\/6167\/revisions\/6170"}],"wp:attachment":[{"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6167"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6167"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/walkingrandomly.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6167"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}