Skip to Main Content

Breadcrumb

Question and Answer

Tom Kyte

Thanks for the question, Alexander.

Asked: October 16, 2007 - 10:25 am UTC

Last updated: October 23, 2007 - 6:46 pm UTC

Version: 10.2.0

Viewed 1000+ times

You Asked

Hello,

We are trying out two POC's for a new environment, both with NAS storage. We experiencing some problems getting the databases configured with NAS. (I should mention the DBAs do not want NAS, we want SAN like everything else but someone is trying to save money)

The situation we are facing is we are either seeing poor I/O performance, or acceptable performance but those options cause the database not to start.

So at this point I am questioning if our tests are valid, and I am looking to you on a good way to test physical I/O from a database point of view (because Oracle caches many things).

One of the things we are running is selecting a bunch of data (1 million rows) and teeing it into a file.

We can't just set the memory to 0, of course Oracle needs memory to start.

What would be your approach? Thanks a lot.

and Tom said...

well, even if you set the SGA to zero, there are caches elsewhere involved here.

The NAS will have a cache - so even if you "alter system flush buffer_cache" (which you can), the NAS caching will be a factor.

In 11g, I have a 'better answer'

http://docs.oracle.com/cd/B28359_01/appdev.111/b28419/d_resmgr.htm#sthref7003


so, you can flush the cache, however.... there are probably other caches out there lurking around!

Rating

  (5 ratings)

Is this answer out of date? If it is, please let us know via a Comment

Comments

Oracle and storage

Philip Papadopoulos, October 17, 2007 - 10:43 am UTC

Check out Kevin's blog. He discusses Oracle and all things storage in great detail.

http://kevinclosson.wordpress.com/

A reader, October 17, 2007 - 11:26 am UTC

I would not use Oracle for storage benchmarking. Use a tool like iozone that is built specifically for this task.

Please explain the following statement: or acceptable performance but those options cause the database not to start.

How are you measuring the "acceptable performance"?

A reader, October 19, 2007 - 4:59 pm UTC

Here is a relevant article that just came out:

How useful are storage benchmarks?

http://searchstorage.techtarget.com/magazineFeature/0,296894,sid5_gci1276093,00.html

NAS choice

A reader, October 22, 2007 - 2:30 am UTC

Hi,

NAS is not a worst choice for a database location!
Todays NAS devices provide sufficient performance with many benefits of management vs. SAN even for a DBA and sufficient performance.

There are several tool to measure performance.
Generally, throughput measured by IOZONE (as mentioned above):
http://www.iozone.org/
Response time and stability by SPECsfs.
http://www.spec.org/sfs93/
For Oracle various IO, there is tool ORASIM:
http://oss.oracle.com/~wcoekaer/orasim/

Tom Kyte
October 23, 2007 - 6:46 pm UTC

I agree, we use lots of NAS ourselves.

ORION may be useful

Charlie B., October 22, 2007 - 10:14 am UTC

OTN has a tool called ORION (Oracle I/O Numbers Calibration Tool) which is available for many platforms. It may be of use too. When I was looking for one, another helpful DBA monitoring Tom's excellent site pointed me to it.
more

Connor and Chris don't just spend all day on AskTOM. You can also catch regular content via Connor's blog and Chris's blog. Or if video is more your thing, check out Connor's channel and Chris's channel from their Youtube channels. And of course, keep up to date with AskTOM via the official twitter account.