Re: Partitioning vs. Disk striping

"Ron Boggs" <rboggs_at_endian-networks.com> wrote in message news:e53f59be.0304081246.58f829e4_at_posting.google.com...
> This is an old debate, but I thought I would resurrect it since I'm
> just getting back into Oracle admin.
>
> An application we are working with is extremely insert intensive...few
> updates or reads in general. Currently, ALL data and indexes are on a
> single 6-disk array with interleaved. (EMC Clarion).
>
> When running peak transactions, the disk performance metrics don't
> look particularly bad except for service times. The array with data
> and indexes show busy less than 50% of the time, but the average
> service time soars into the 100's. During normal opertion, the
> service times are in the 20-30 range. Accepted that it will increase
> with load, but this is non-linear. The CPU metrics during peak use
> are fine...50-60% used.
>
> My question is, in 'general', is striping the data and indexes across
> a 6 disk array enough to prevent bottlenecks at the disk or is it
> better to use Oracle partitioning to address this? Is one solution
> better at mixed read/write vs. 'nearly' all writes?

Hi Ron,

My comments here is you are discussing two entirely different things that are not mutually exclusive by any means.

By "striping" your data appropriately, you can improve the effectiveness of retrieving the data required by Oracle by reducing contention on a single disk, by increasing I/O throughput, by enabling more effective use of parallelism, etc.

However by using partitioning appropriately, you can reduce the amount of data that needs to be retrieved or manipulated in the first place.

So striping enables "more workers to dig the hole", while partitioning can "reduce the size of the hole" (I'm sorry it's late and it's the best analogy I can come up with :)

So both have their uses and it's kinda incorrect to have them compared to each other per se. I would suggest partitioning and striping of the (sub)partitions as appropriate.

Cheers

Richard Received on Wed Apr 09 2003 - 08:53:01 CDT