Re: Parallel Server and availability?

Uli Harder wrote in message <36B08615.E4851BC4_at_noether.demon.co.uk>...

>However this comment confuses the issue: There are two things

>a) Oracle Parallel Query Option (PQO) which can be run either on single
>instance databases or Oracle Parallel Server. All it does is splitting
>up full table scans over several processors.

Yes.

>b) Oracle parallel server CAN be run without PQO (AFAIK) and in a way
>has got nothing to do with PQO.

OPS with PQ is as useless as a car without weels IMHO. Sure you can drive a car without tires on it, BUT WHY? :-)

OPS without PQ simply gives you multiple instances on the same database. It thus spread the process load (which can be done better and cheaper with SMP IMHO), and the only added benefit is that you have high availibility - which still means -nothing- to the user who will lose all his processes on an OPS node if it crashes. And at what expense?

>OPS does come with all sorts of failover
>and recovery options depending on hardware and operating system.

Any idea how good NT is in this department?

>main problem with its implementation is the partitioning of data to
>avoid contention in the PCM lock management.

I would rather say that the major problem differs depending on what you're trying to do. I never had any PCM lock problems. But then this was a data warehouse and not an OLTP installation.

>Anyway, I just wanted to make sure no one confuses the Oracle parallel
>query option with Oracle Parallel Server.

Thanks for the comments. But I still think that PQ goes hand in hand with OPS. The only place I can see OPS working without PQ is for certain types of OLTP systems. But in such a case, having OPS only providing database high availbility, is not cost justified IMHO.

regards,
Billy Received on Fri Jan 29 1999 - 03:35:15 CST