Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Mailing Lists -> Oracle-L -> RE: 10G and UFS - long write times

RE: 10G and UFS - long write times

From: Hallas, John, Tech Dev <>
Date: Thu, 8 Jun 2006 14:42:52 +0100
Message-ID: <1C6E45ADB2EC324F9553E468ABFE0F63034BDD9D@UKWMXM04>

I would be very keen if others can run the test specified in my earlier post and compare 9i and 10g on UFS mounted disk, on Solaris would be even better.

The test is very simple but whilst we are a very big site I do not have that many 10G databases on local disk setups to really prove the case either way.

From a few responses I have I think people are confusing what I am actually saying. I am not expecting UFS to be as fast as SAN. However I am expecting 9i to be the same as 10g on UFS and that is not what I am seeing. I am looking at 10g being twice as slow as 9i in the scenario I have provided which whilst maybe not realistic is a valid test. That is write and commit 50,000 times

I have not created the 10G databases any differently than I would the 9i ones. They essentially have many of the same init.ora parameters. If we need to tweak specifically for 10G on UFS then so be it but I need to fully understand the issue first


-----Original Message-----
From: Mladen Gogala [] Sent: 08 June 2006 14:29
To: Hallas, John, Tech Dev
Cc: Ron Rogers;; Subject: Re: 10G and UFS - long write times

On 06/08/2006 09:06:33 AM, Hallas, John, Tech Dev wrote:
> More testing has confirmed that the issue is there in 9i but not to
> same degree
> Very simply, what we are seeing is that using the test routine I
> earlier on UFS mounted local disk the results are :-
> 9i SAN time = (15-30s) * 2.5 = 3.30 minutes
> 10g San time = ((15-30s) * 14 = 6 - 9 minutes
> This is on Solaris and seems to me to be repeatable on Linux but the
> Unix boys want me to stay with Solaris for now for ease of testing.
> John

This is extremely worrisome to me. The company that I work for has a bunch of
9i databases and a sandbox to test 10G. As we are primarily DW place, writes weren't
tested yet. How many people here have implemented 10GR2 in production and is there
a consensus about the problem with the speed of writing to the database? I'll search
Metalink later.

Mladen Gogala

Received on Thu Jun 08 2006 - 08:42:52 CDT

Original text of this message