Re: Cosign in production at UoA and thanks

[David Alexander <alexandd@xxxxxxxx>]

--On Tuesday, June 14, 2005 10:31 AM -0400 Wesley Craig <wes@xxxxxxxxx> 
wrote:

> On 14 Jun 2005, at 10:06, David Alexander wrote:
> > We implemented Cosign behind a F5 BigIP load balancer.  Our short-
> > term solution uses NFS as a shared file store between blades.
> > Although we have a high availability architecture for NFS, we are
> > concerned with the response time of NFS under load.  In the future,
> > we would like to replace the file reads and writes with Oracle
> > database calls.
>
> So you're not using replication at all?  Can you describe how you've
> deployed NFS with high-availability?

We are not using Cosign's replication.  We configured each Cosign blade to 
store cookies on the NFS mount point.

NFS is hosted on an Alpha cluster with memory channel interconnect.  NFS 
can fail over between cluster members.  We did a lot of performance testing 
of this NFS system because we are using it in production for other 
applications.  The disk use profile of Cosign - lots of small reads and 
writes - does not cause contention on the NFS cluster.  The NFS solution is 
a temporary kludge until we can get Cosign working with Oracle.