Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Several CPUs and the parallel option
Hi Roobaron,
If a single CPU is waiting for a *serial* operation to complete (ie. processing the blocks in a multiblock read), then it would appear that several process working in *parallel* processing multiple blocks *simultaneously* would go kinda quicker.
Compare a nonparallel update of 1,000,000 rows in a partitioned table (non partitioned in 9i) on one CPU with heaps in reserve vs a suitably tuned parallel operation (assuming good I/O distribution and so minimal contention) and I would love to put a fiver on the winner.
Regards
Richard
"roobaron" <member_at_dbforums.com> wrote in message
news:3d44de9a$1_at_news.usenetgateway.com...
> Is your system CPU bound?
>
> Your app will only benefit from using extra CPU if it is CPU bound.
>
> To deflect or nullify the blame game you mention in the last
> paragraph... read some documents about the Oracle Wait interface. Search
> google.com for Oracle event 10046.
>
> You can then trace a session in the database and see what it is waiting
> on. If it is CPU then proceed with fixing it, if it is IO fix it first
> before looking at CPU.
>
> Have a look at the forum threads via searching, to see what other people
> have asked about tracing etc.
>
> I have found that applications which do sorting or large table scans can
> benefit from parallel query servers (see other answer for more details)
> and also tables which have been partitioned will use parallel query
> servers to retrieve data if that is the best method.
>
> Hope this helps
>
> Have Fun
>
>
>
> --
> An Oracle DBAs undocumented init.ora parameters
> _OVERCOME_POOR_DESIGN = TRUE
> _DISABLE_BAD_PERFORMANCE = TRUE
>
> Posted via dBforums
> http://dbforums.com
Received on Mon Jul 29 2002 - 02:37:00 CDT
![]() |
![]() |