{"id":264,"date":"2011-08-18T08:59:10","date_gmt":"2011-08-18T14:59:10","guid":{"rendered":"http:\/\/www.unifiedcomputingblog.com\/?p=264"},"modified":"2011-08-18T08:59:10","modified_gmt":"2011-08-18T14:59:10","slug":"fcoe-vs-iscsi-vs-nfs","status":"publish","type":"post","link":"http:\/\/www.unifiedcomputingblog.com\/?p=264","title":{"rendered":"FCoE vs. iSCSI vs. NFS"},"content":{"rendered":"<p>The following was just a short note I wrote in an internal discussion about FCoE vs. iSCSI vs. NFS &#8211; and spurred by <a href=\"http:\/\/datacenteroverlords.com\/2011\/08\/16\/jinkies-its-an-fcoe-mystery\/\" target=\"_blank\">Tony Bourke&#8217;s discussion<\/a> about methods for implementing FCoE.<\/p>\n<p>This wasn&#8217;t intended to be a detailed analysis, just a couple of random musings.\u00a0\u00a0 Comments as always are welcome.<\/p>\n<p>&#8212;&#8211;<\/p>\n<p>While NFS and iSCSI are completely different approaches to accessing<br \/>\nstorage, they both &#8220;suffer&#8221; from the same ailment &#8211; TCP.\u00a0\u00a0 Remember folks,<br \/>\nTCP was developed in the 70&#8217;s for the express purpose of connecting<br \/>\ndisparate networks over long, latent, and likely unreliable links.\u00a0 The<br \/>\noverheads placed onto communication solely to address these criteria<br \/>\nsimply aren&#8217;t appropriate in the datacenter.\u00a0 We&#8217;re talking about a<br \/>\nprotocol written to support links slower than your Bluetooth headset.\u00a0 \ud83d\ude42<\/p>\n<p>iSCSI is a hack, plain and simple.\u00a0 It solves a cost problem, not a<br \/>\ntechnology one.\u00a0 Even its name is misleading &#8211; iSCSI.\u00a0\u00a0 It isn&#8217;t SCSI over<br \/>\nIP &#8211; it&#8217;s SCSI over TCP over IP. So call it tSCSI or tiSCSI.<\/p>\n<p>I&#8217;m not saying they&#8217;re not &#8220;good enough&#8221;, but why do &#8220;good enough&#8221; now<br \/>\nthat &#8220;better&#8221; is getting much closer in price?\u00a0 On the array side, I<br \/>\nexpect more and more vendors to go the NetApp route &#8211; all protocols in one<br \/>\nbox &#8211; just turn on which ones you want to use (via appropriate licensing,<br \/>\nof course).\u00a0 10G DCB makes this even easier and more attractive &#8211; one<br \/>\nport, you pick the protocol you&#8217;re comfortable with.<\/p>\n<p>As one of my coworkers points out, FCoE is a bit of a cannon &#8211; and for many customers,<br \/>\ntheir storage challenges are more in mosquito scale.<\/p>\n<p>Fibre Channel was developed with storage in mind as a datacenter protocol,<br \/>\nI haven&#8217;t seen one yet I like better for moving SCSI commands around *in<br \/>\nthe datacenter*.\u00a0\u00a0 I&#8217;m sure someone will develop a new protocol at some<br \/>\npoint that utilizes DCB-specific architectures to replace iSCSI and<br \/>\nFCoE&#8230; but why?\u00a0\u00a0 If you want a high performance, low latency,<br \/>\nmade-for-storage protocol, run FC over whatever wire you feel like.\u00a0\u00a0 If<br \/>\nyou want a low-cost solution utilizing commodity<br \/>\nhardware\/switching\/routing, use iSCSI and\/or NFS.\u00a0\u00a0 I don&#8217;t know that<br \/>\nthere&#8217;s a new problem to solve here.<\/p>\n<p>For customers that already have and know FC, FCoE is a no-brainer.<br \/>\nNothing new to learn about how to control access, you&#8217;re just replacing<br \/>\nthe wires.\u00a0 iSCSI and NFS introduce whole new mechanisms and mindsets into<br \/>\naccessing storage if you&#8217;re not used to them.<\/p>\n<p>I saw a quote the other day that said that Fibre Channel is like smoking &#8211;<br \/>\nif you&#8217;re not already doing it, there&#8217;s no reason to start now.\u00a0\u00a0 I get<br \/>\nthe sentiment, but I don&#8217;t agree.\u00a0\u00a0 FC as a protocol is the right tool for<br \/>\na lot of jobs &#8211; but it&#8217;s not the right tool for every job.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>The following was just a short note I wrote in an internal discussion about FCoE vs. iSCSI vs. NFS &#8211; and spurred by Tony Bourke&#8217;s discussion about methods for implementing FCoE. This wasn&#8217;t intended to be a detailed analysis, just a couple of random musings.\u00a0\u00a0 Comments as always are welcome. &#8212;&#8211; While NFS and iSCSI &hellip; <a href=\"http:\/\/www.unifiedcomputingblog.com\/?p=264\" class=\"more-link\">Continue reading <span class=\"screen-reader-text\">FCoE vs. iSCSI vs. NFS<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[5],"tags":[23,32,40],"class_list":["post-264","post","type-post","status-publish","format-standard","hentry","category-miscellaneous","tag-fcoe","tag-iscsi","tag-nfs"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=\/wp\/v2\/posts\/264","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=264"}],"version-history":[{"count":0,"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=\/wp\/v2\/posts\/264\/revisions"}],"wp:attachment":[{"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=264"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=264"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.unifiedcomputingblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=264"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}