Attachment 47028 Details for Bug 75341 – Gentoo Cluster File System Selection Guide

Gentoo Cluster File System Selection Guide

cluster-fs-guide.xml (text/plain), 3.70 KB, created by Sven Vermeulen (RETIRED) on 2004-12-28 07:25:51 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Sven Vermeulen (RETIRED)

Created: 2004-12-28 07:25:51 UTC

Size: 3.70 KB

patch

obsolete

><?xml version='1.0' encoding='UTF-8'?>
>
>
>
><!DOCTYPE guide SYSTEM "/dtd/guide.dtd">
>
><guide link="cluster-fs-guide.xml">
><title>Gentoo Cluster File System Selection Guide</title>
>
><author title="Author">
>  <mail link="iggy@gentoo.org">Brian Jackson</mail>
></author>
>
><abstract>
>Clusters benefit from advanced file systems. This guide offers a quick run-down
>on the possible file systems available.
></abstract>
>
>
>
><license/>
>
><version>1.0</version>
><date>2004-12-28</date>
>
><chapter>
><title>Selection of a File System for a Cluster</title>
><section>
><title>Introduction</title>
><body>
>
><p>
>For purposes of this document we will assume there are 2 types of clusters. 
></p>
>
><ul>
>  <li>High Availability (HA) clusters</li>
>  <li>High Performance Computing (HPC) clusters</li>
></ul>
>
><p>
>To select the proper fs it is important to know the features of each filesystem
>and the features that you need out of a filesystem. 
></p>
>
></body>
></section>
><section>
><title>NFS</title>
><body>
>
><p>
>Classic network filesystem.
></p>
>
><table>
><tr>
>  <th>Advantages</th>
>  <th>Disadvantages</th>
></tr>
><tr>
>  <ti>Very stable</ti>
>  <ti>Many SPOFs</ti>
></tr>
><tr>
>  <ti>Widely used</ti>
>  <ti></ti>
></tr>
></table>
>
></body>
></section>
><section>
><title>OpenGFS</title>
><body>
>
><p>
>multimaster read/write shared storage filesystem.
></p>
>
><table>
><tr>
>  <th>Advantages</th>
>  <th>Disadvantages</th>
></tr>
><tr>
>  <ti>No single master</ti>
>  <ti>Still has a SPOF</ti>
></tr>
></table>
>
></body>
></section>
><section>
><title>Oracle Cluster File System</title>
><body>
>
><p>
>shared storage filesystem 
></p>
>
><table>
><tr>
>  <th>Advantages</th>
>  <th>Disadvantages</th>
></tr>
><tr>
>  <ti>Strong commercial backing</ti>
>  <ti>Not useful as general purpose file system yet</ti>
></tr>
></table>
>
></body>
></section>
><section>
><title>Lustre</title>
><body>
>
><p>
>Lustre is a novel storage and file system architecture and implementation
>suitable for very large clusters.
></p>
>
><table>
><tr>
>  <th>Advantages</th>
>  <th>Disadvantages</th>
></tr>
><tr>
>  <ti>Well suited for very large clusters</ti>
>  <ti>Still quite new</ti>
></tr>
><tr>
>  <ti></ti>
>  <ti>Not all redundancy features are implemented yet</ti>
></tr>
></table>
>
></body>
></section>
><section>
><title>Coda</title>
><body>
>
><p>
>Advanced network filesystem with origins in AFS2 
></p>
>
><table>
><tr>
>  <th>Advantages</th>
>  <th>Disadvantages</th>
></tr>
><tr>
>  <ti>Client side caching (i.e. disconnected operation)</ti>
>  <ti>it's crap</ti>
></tr>
><tr>
>  <ti>Server replication</ti>
>  <ti></ti>
></tr>
></table>
>
></body>
></section>
><section>
><title>Intermezzo</title>
><body>
>
><p>
>Very similar feature-wise to coda.
></p>
>
></body>
></section>
></chapter>
><chapter>
><title>What File System to Choose?</title>
><section>
><body>
>
><p>
>NFS is probably the best choice for most cluster's just because of it's
>pervasiveness, stability, and relative speed.
></p>
>
><p>
>For a high availability cluster you can either use a regular single node fs
>(reiserfs,ext3,etc.) that is only mounted on a single node at a time.
></p>
>
><p>
>For HPC clusters lustre or some mixture of OpenGFS and NFS is a good choice
></p>
>
><p>
>For those inbetween type clusters (lvs, etc.) NFS is a good choice for most,
>but you could also use any of the above or a mixture dpending on your
>requirements.
></p>
>
><p>
>Most people are aware that NFS has no type of redundancy built into it, but
>there are things you can do to make a single NFS box more highly available. Use
>redundant power supplies, some redundant level of RAID, quality hardware,
>traditionaly speaking scsi drives have been better suited to round the clock
>operation, etc. 
></p>
>
></body>
></section>
></chapter>
>
></guide>

Actions: View

Attachments on bug 75341: 47028 | 47029 | 47030