John,
I believe that as of Lustre 1.4.0 (or a late 1.2.x) the OSTs no longer
perform any sort of caching of data. Therefore if you have multiple
clients reading the same file then each request will have to go to disk
to be satisfied.
Fergal.
--
Fergal.McCarthy@HP.com
(The contents of this message and any attachments to it are confidential
and may be legally privileged. If you have received this message in
error you should delete it from your system immediately and advise the
sender. To any recipient of this message within HP, unless otherwise
stated, you should consider this message and attachments as "HP
CONFIDENTIAL".)
-----Original Message-----
From: lustre-discuss-admin@lists.clusterfs.com
[mailto:lustre-discuss-admin@lists.clusterfs.com] On Behalf Of John R.
Dunning
Sent: 10 February 2006 15:44
To: lustre-discuss@lists.clusterfs.com
Subject: [Lustre-discuss] Caching at the OST/LOV
Question for the experts: Is there a description of lustre''s overall
caching strategy? I''m particularly interested in what kind of caching
happens at the OST/LOV level, and how I can control it, so as to tune
what happens when I have large numbers of clients asking for common
subsets of data.
I''ve re-read the white papers and some of the other docs, but perhaps
I missed this part?
Also, one doc I read said there was ongoing work on a collaborative
cache, allowing clients to fetch stuff from other clients. Is that
still ongoing, and if so, what''s its state?
TIA...
_______________________________________________
Lustre-discuss mailing list
Lustre-discuss@lists.clusterfs.com
https://lists.clusterfs.com/mailman/listinfo/lustre-discuss