1 This topic is a place to collect information on AFS performance measurements and tools.
5 <li><a href="#Tools"> Tools</a><ul>
6 <li><a href="#Andrew Benchmark"> Andrew Benchmark</a></li>
7 <li><a href="#afsfsperf (low-level fileserver"> afsfsperf (low-level fileserver performance tester)</a></li>
8 <li><a href="#PostMark"> PostMark</a></li>
11 <li><a href="#Measurements"> Measurements</a><ul>
12 <li><a href="#AFS vs NFS"> AFS vs NFS</a></li>
13 <li><a href="#Afsfsperf collection"> Afsfsperf collection</a></li>
14 <li><a href="#Solaris x86 Inode vs Red Hat Nam"> Solaris x86 Inode vs Red Hat Namei on same hardware</a></li>
22 ## <a name="Tools"></a> Tools
24 ### <a name="Andrew Benchmark"></a> Andrew Benchmark
26 This old and venerable suite has been used many times. It is mentioned in the [FAQ](https://grand.central.org/twiki/bin/view/AFSLore/AdminFAQ#3_18_What_is_the_Andrew_Benchmar).
30 ### <a name="afsfsperf (low-level fileserver"></a><a name="afsfsperf (low-level fileserver "></a> afsfsperf (low-level fileserver performance tester)
32 [[ArlaAFS]] comes with a tool called afsfsperf, apparently available as part of their release. afsfsperf is a small programm that acts like an minimal afs-client to observe the performace of the fileserver. For relevant numbers you need to run your instance of afsfsperf on a machine that is at least as fast as the fileserver.
36 ### <a name="PostMark"></a><a name="_PostMark"></a> PostMark
38 A [NetApp paper on PostMark](http://www.netapp.com/tech_library/3022.html).
40 The numbers below were made with version 1.53 of PostMark from [[NetApp]]. <br /> Older versions of PostMark output wrong results for the number of "files deleted"!!<br /> Here is version [1\.53 of PostMark](http://www.slac.stanford.edu/~alfw/postmark-1.53.c.gz) [Jeffrey Katcher](mailto:katcher@netapp.com) sent me.
44 ## <a name="Measurements"></a> Measurements
46 ### <a name="AFS vs NFS"></a> AFS vs NFS
48 A chart showing scalability is in the [FAQ](https://grand.central.org/twiki/bin/view/AFSLore/GeneralFAQ#1_11_How_does_AFS_compare_with_N).
52 ### <a name="Afsfsperf collection"></a> Afsfsperf collection
54 A large collection of information. All numbers done with afsfsperf from the [[ArlaAFS]] package. Varios of OS'es and filesystems benchmarked.
56 <http://www.e.kth.se/~jimmy/afsfsperf/afsfsperf.html>
60 ### <a name="Solaris x86 Inode vs Red Hat Nam"></a> Solaris x86 Inode vs Red Hat Namei on same hardware
62 From a [message](https://lists.openafs.org/pipermail/openafs-info/2003-April/008916.html) on the [[OpenAFS-Info]] mailing list:
64 > I was wondering how Solaris x86 would be doing as AFS fileserver compared to Red Hat 7.3 Linux on the same hardware.
66 > I've conducted some benchmarks using [[PostMark|Main/PerformanceWork#PostMark]] and arla's afsfsperf to get an answer (not _the_ answer!) to this question.
68 > I've used [[OpenAFS]]-1.2.8 with inode and namei interface on Solaris with non-logging and logging file system for the /vicepa partition.
70 > On Linux I've used [[OpenAFS]]-1.2.8 (namei only) with ext2, ext3, reiserfs v1, and reiserfs v2 for /vicepa.
72 > The results are available under [http://www.slac.stanford.edu/~alfw/AFS\_bench2.html](http://www.slac.stanford.edu/~alfw/AFS_bench2.html)
74 > Alf Wachsmann | e-mail: alfw@slac.stanford.edu
75 > SLAC Computing Service | Phone: +1-650-926-4802
76 > 2575 Sand Hill Road, M/S 97 | FAX: +1-650-926-3329
77 > Menlo Park, CA 94025, USA | Office: Bldg. 50/323
78 > http://www.slac.stanford.edu/~alfw (PGP)
82 -- [[TedAnderson]] - 25 Apr 2003