[OpenAFS] File systems on Linux, again.

chas williams - CONTRACTOR chas@cmf.nrl.navy.mil
Fri, 30 Nov 2007 12:58:18 -0500

no i dont think rx locking is the problem.  the rx locking is
actually pretty good.  i had tracked this down with fstrace
at one point but i seem to have lost the trace at the moment.
i will dig around and see if i can find it.

In message <475043F7.1010101@linuxbox.com>,Matt Benjamin writes:
>Hash: SHA256
>Hey, Chas,
>Sorry to bug.
>I've been looking at this, tangentially, because I've been working with
>bypassing dcache/memcache, writing direct into page cache.  Pretty far
>along on that side of things.  Is rx locking so coarse that in general
>only one read makes progress even independent of the cache--to your
>chas williams - CONTRACTOR wrote:
>> In message <E2E34866B20B824DA894CA53D1CF0DE204C9E912@ehost005-3.exch005inter
>> ia.net>,"Jerry Normandin" writes:
>>> write performance is actually impressive.  file creation and deletion
>>> are very slow on afs.
>> because writing is easier than reading.  the afs cache manager can
>> group the outgoing writes together and send them in a single message.
>> while the cache manager has readahead it doesnt work because the afs
>> global locks blocks any progress the readahead thread might make.
>> _______________________________________________
>> OpenAFS-info mailing list
>> OpenAFS-info@openafs.org
>> https://lists.openafs.org/mailman/listinfo/openafs-info
>- --
>Matt Benjamin
>The Linux Box
>206 South Fifth Ave. Suite 150
>Ann Arbor, MI  48104
>tel. 734-761-4689
>fax. 734-769-8938
>cel. 734-216-5309
>Version: GnuPG v1.4.7 (GNU/Linux)
>Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>OpenAFS-info mailing list