On Thu, 16 May 2002, Chris Marget wrote: > > Actually, Oracle has officially endorsed using Netapp Filers with Oracle > > 9i, and both of them seem to claim that this is a direction that they want > > to encourage the user base to go in, because to the way that the snap > > technology can simplify hot backups on large databases. Who would have
It's not just hot backups, you can also abuse snapshots to run a read-only Oracle instance for data warehouse/report crunching purposes. http://www.netapp.com/tech_library/3041.html We don't have that working yet -- the PeopleSoft process scheduler needs to be able to write to at least some tablespaces -- but are very interested. You absolutely could not do that with a copy-on-write snapshot implementation like EMC's. You would need to make another copy of the data. WAFL doesn't have this problem. > i haven't done the math, but i can't imagine that any nfs version over > any network medium can even begin to approach the performance of raw > disk access over disparate HBAs... Proper gig fiber channel would clearly perform better, but GigE isn't that bad. The long-term answer is replacing NFS with DAFS/VI. NetApp demonstrated that nearly a year ago, but it's not near production yet. We do pretty well with two dozen Oracle databases on 3 Suns connected to a NetApp 820. There was a period of tuning to make sure frequently abused tables get pegged in RAM, but the Suns are all CPU-bound now. It helps that our average database is only 6GB. The NetApp is essentially a reliable 1.5GB network-attached RAM disk most of the time. We did have one major unrecoverable Oracle error that could be blamed on NAS versus SAN or local disk. I broke the NetApp's /etc/exports on a live database once. Oracle has no problem doing instance recover from snapshot or in case of network jitter, but if some processes can continue write across existing NFS handles and others can't write to the same filesystem, it can choke in pathological cases. Fortunately this was just a test/migration database. You do want to duplex redo and archive logs on both netapp and local disk just in case. -- Rich Graves <[EMAIL PROTECTED]> UNet Systems Administrator --- Send mail for the `bblisa' mailing list to `[EMAIL PROTECTED]'. Mail administrative requests to `[EMAIL PROTECTED]'.
