SQL Clustering requires a storage medum that is physically connected to all
the host servers in a cluster. The Cluster service arbitrates actual
ownership of the disks to avoid data corruption issues. Here is a very good
introduction and overview on SQL failover clustering:
SQL Server 2000 Failover Clustering
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/failclus.mspx

Signature
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
> Hello,
> We are looking to implement (finally) some redundancy into our SQL Server
[quoted text clipped - 6 lines]
> Thanks,
> John
[SEE inline]

Signature
Hope it helps
dw
_______________________________
Don Wilwol
donwilwol(DELETE)@yahoo.com
http://spaces.msn.com/members/wilwol/
> Hello,
> We are looking to implement (finally) some redundancy into our SQL Server
> setup. We are interested in failover clustering, but from what we've seen
> so far most people use a SAN to implement it.
You are seeing scsi attached clusters less and less as SAN prices come down,
but there are still plenty of them out there. Its still a viable solution.
We'd prefer using independent
> servers, each with their own RAID array (similar to replication, but with
> the seamless failover capabilities of clustering). Is this scenario
> possible or recommended?
There are third part applications that make this possible. I'm in the
process of testing NSI's Double Take and Geo cluster applications right now.
We hope to use it to set up a DR site for both SQL and Exchange. I've also
heard of one called neverfail that I haven't looked into, but may just to
see what its like.
> Thanks,
> John
John - 31 Aug 2005 14:09 GMT
Thanks for your input. From what I have seen of SANs, they simply cannot
beat SCSI RAID in terms of performance for the dollar. The DB we're looking
at is running on a 16 x 15,000 rpm raid 10, with a 4 x 10000rpm raid 10 for
logs. SANs that can approach that level of performance all appear to be
well over 25,000, which is too high of an investement for our company to
implement failover clustering at this time. It looks like we will go with
log shipping or perhaps transactional replication.
Thanks!
> [SEE inline]
>
[quoted text clipped - 19 lines]
> > Thanks,
> > John