[Lustre-discuss] Lustre reading file and RAM

Mag Gam magawake at gmail.com
Sat Jan 2 06:40:47 PST 2010


Thats what I am doing now. There must be cache on the MDS side on 1.6, no?


On Sat, Jan 2, 2010 at 1:06 AM, Andreas Dilger <adilger at sun.com> wrote:
> On 2010-01-01, at 15:38, Mag Gam wrote:
>>
>> We are running 1.6.7.2 OSS and Clients. If there are 50 clients
>> accessing the same 10gig file would the 10gig file go into memory of
>> the OSS or for each access it starts over? My OSS has 32gig of memory.
>
>
> The OSS read-only cache feature was not introduced until 1.8.  With 1.6 it
> will read the same data from disk each time.
>
> Note that if you have multiple OSS nodes it makes sense to stripe the file
> over all of the OSS nodes, since it is large and many clients access it, it
> will increase the aggregate bandwidth for access to that file.
>
> Cheers, Andreas
> --
> Andreas Dilger
> Sr. Staff Engineer, Lustre Group
> Sun Microsystems of Canada, Inc.
>
>



More information about the lustre-discuss mailing list