latest/greatest oracle OPS (oracle parallel server)

[Michael Murphy_2]

Anyone out there have experience with Oracle Parallel Server (OPS)? We currently have version 8.1.6 on another platform - pretty happy with it (sometime hangs when interconnect is interrupted). Looking for experiences with the newest version (rac?) - happy with it? Any other ways to go for high availability? We are planning an implementing two node on n-class...

[Hein van den Heuvel · Accepted solution]

OPS was a lot about Parallel processing, and a little about high-availabilty.
Your generic High Availability solution was a lot about Fail-Over (Hot - Cold, Active - Standy).
RAC tries to be the best of both worlds.
Highly Available, active-active, Parallel.
Better then Hot-Cold ( 1 out of 2 ) it can give you N - out of - M ( eg 3 out of 4 ) availability.
So if you size a solution to run say 60% load over 4 servers (classic SAP suggestion) and one would fail, the 3 others could readily deliver 80% each and the users might not notice performance degradation!

Oracle made great effort, and progress, to elimiate classical OPS bottlenecks. Cache - Cache data block shipping instead of pinging over the disk. Better lock (granularity) managent.
There is some of course hype there, but a lot of goodness also. Well worth checkig out.

A lot has been written on the subject. Please search and read the Oracle Doc, this forum, Oracle metalink, Oracle Openworld presentations for lots of info!

It would be good to hear some actual usage feedback.
Myself, I worked on pretty succesfull 3-tier SAP - Oracle RAC benchmark on Tru64. We had a dedicated 'memory channel' for RAC communication there. That helped making the succes. Other platforms use Gigabit connections.

hth,
Hein.