[Gluster-devel] running RDBMS on top of GlusterFS

[Mike Machuidel]

Of course. When indexes are used and queries are well optimized there
will be far less disk activity. You may also tune the caching
parameters of your RDBMS which will improve performance and lower te
disk activity as well. Also make sure the kernel is properly tuned for
database cases (ex. increasing the shared memory segment size).

You might like to read the following texts (aimed at postgresql):
http://developer.postgresql.org/pgdocs/postgres/kernel-resources.html
http://momjian.us/main/writings/pgsql/hw_performance/index.html

On 6/19/07, Marcin Krol <admin at domeny.pl> wrote:
> Mike Machuidel napisał(a):
> > Hi Marcin,
> >
> > I think you should put the RDBMS in the storage tier (the same level /
> > layer as the bricks). By doing this only queries and their results
> > will go over the network instead of entire database files (which can
> > be very slow and dangerous). If you want redundancy you should then
> > use the RDBMS its replication feature.
> That was my first thought - however, we may have lots of spare CPU
> cycles on client machines (Node A and B in application layer in your
> example picture), and for peculiar reasons little CPU time to spare on
> the bricks, which will handle some other important functions as well.
>
> There are situations in which logical I/O in databases may account for a
> lot of CPU time and not that much disk activity. Hence the question.
>
> >
> > See the following for an example:
> > http://www.satl.com/~machuidel/glusterfs/2tier_gluster.png
> >
> > Mike Machuidel
> >
> > On 6/19/07, Marcin Krol <admin at domeny.pl> wrote:
> >> Hello everyone,
> >>
> >> First, thanks to developers for GlusterFS - we are considering
> >> transition to cluster filesystem and so far I have narrowed the
> >> possible choices down to only two usable options, namely GlusterFS and
> >> LUSTRE.
> >>
> >> I have searched archives for issues related to running SQL engine on
> >> top of GlusterFS, but haven't found explicit answers, so here it
> >> goes:
> >>
> >> 1. Are there any drawbacks to running RDBMS such as PostgreSQL or
> >> MySQL on client node while the data like .MYI and .MYD files are
> >> stored on GlusterFS brick?
> >>
> >> Conversely, are there any advantages? E.g. any thoughts about
> >> configuration that could parallelize reads of different sections of
> >> index and data files and thus boost performance of SELECT queries?
> >>
> >> 2. Any hard data / tests / hints about performance in comparison to
> >> running RDBMS engine on local fs?
> >>
> >> 3. Can you think of any problems regarding reliability, locking
> >> issues, etc?
> >>
> >> It would be nice to add an item like this to FAQ - I'd write smth on
> >> the topic myself, but frankly I have very little experience with
> >> cluster filesystems, so it's not like I feel knowledgeable on the
> >> subject.
> >>
> >> --
> >> Marcin Krol
> >>
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> Gluster-devel mailing list
> >> Gluster-devel at nongnu.org
> >> http://lists.nongnu.org/mailman/listinfo/gluster-devel
> >>
> >
>
>
> --
> Dział Techniczny
> Marcin Król
>
> Domeny, Hosting, Kolokacja, Certyfikaty SSL, Monitoring serwerów ...
> ------------------------------------------------------------------
> DOMENY.PL sp. z o.o. ul. Wielicka 50, 30-552 Kraków, Poland
> tel. +48(12)296 3663, info: +48 501 DOMENY
> fax. +48(12)296 3664, +48(22)3 987 365
> e-mail: pomoc at Domeny.pl, www: http://www.Domeny.pl
> ------------------------------------------------------------------
> Komunikator online/ Live Chat:
> http://live2.domeny.pl/request_email.php?l=phplive&x=1&deptid=2
>
>
>