[Gluster-users] GlusterFS as underlying Replicated Disk for App Server

[Dan Hawker]

Well my experience with Oracle is about 8 years out of date so take this
> for what it is.
>
> Trust me, this is Oracle. The old stuff is always still around somewhere
just lurking in the background.


> In the past I've used Oracle with OCFS (Oracle Cluster File System) on a
> shared disk on a SAN.
> In my experience it was always slow because of the safeties it requires
> for multiple instances using a shared disk. So its no surprise the
> performance on NFS wouldn't be great
>
> That said under the hood whenever possible it loads whole files into RAM
> to extract data however it does misplace writes. So it may not perform as
> well on writes as you may hope, but neither will NFS.
>
> The reason why Oracle has started supporting NFS is really to try to
> compete in the cloud market not because its a good idea.
>
> In general distributed Databases using shared disk is a bad idea and
> should be avoided if possible.
>
> Yeah I've used OCFS in the past for a specific area of Oracle DB work
(IIRC they used to recommend using it for the voting disks within RAC).
That was one option with 10g and earlier RAC. That's been deprecated (I
believe in Oracle DB 11gR1 or maybe R2) as they want everyone to use
clusterware and ASM to manage disks, clusterFS's and the like. Which makes
sense because OCFS is the devils work.


> That said consider talking to the people at EnterpriseDB they may have a
> better compatible alternative for you because the optimize the PostgreSQL
> stack to work in the cloud and provide a Oracle client compatibility layer.
> In addition they can easily configure their version of PostgreSQL to do
> more efficient Hot spare (read write on the primary and read-only on the
> backup) replication via the databases journal which will be far more
> efficient as far as speed and your network utilization.
>
>
> We're committed to Oracle DB due to the product set we're using (far, far
too many Oracle apps). We have replication for the DB in place using
DataGuard (Oracle replication tech for DB's) so am not too worried about
that.

This is all about the middleware/application servers that host the apps,
rather than the DB itself. Weblogic has a requirement for a shared
persistence layer in WLS clusters, and we need to provide that. Someone
just bought the wrong NAS head, so am exploring options...

Thanks

Dan


-- 
--
Dan Hawker
--
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20131008/8cd882f4/attachment.html>