Re: Stress my CPU's
- From: Kellyn Pedersen <kjped1313@xxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx, japplewhite@xxxxxxxxxxxxx
- Date: Tue, 17 Nov 2009 08:16:16 -0800 (PST)
I agree with Jack- We were experiencing some serious CPU tension on one of our
mart servers and it turned out to come from one of the app support gzipping
files to an nfs mount. Gzip is a notorious hogger of CPU!
Kellyn Pedersen
Multi-Platform DBA
I-Behavior Inc.
http://www.linkedin.com/in/kellynpedersen
"Go away before I replace you with a very small and efficient shell script..."
--- On Tue, 11/17/09, japplewhite@xxxxxxxxxxxxx <japplewhite@xxxxxxxxxxxxx>
wrote:
From: japplewhite@xxxxxxxxxxxxx <japplewhite@xxxxxxxxxxxxx>
Subject: Re: Stress my CPU's
To: oracle-l@xxxxxxxxxxxxx
Cc: Lee.Robertson@xxxxxxxxxx
Date: Tuesday, November 17, 2009, 9:03 AM
Not a database-generated CPU stress, but I find that gzip / gunzip really suck
up CPU. Sometimes I launch 8 parallel gzip sessions and bring the 8 CPUs on
one of our newer servers to near 100% utilization each.
Write a little script that spawns 16 gzip -9 processes, then gunzips, then
gzips on some multi-GB sized files. Since the CPUs won't care how they're
stressed - database or otherwise - you might get the results you want.
Jack C. Applewhite - Database Administrator
Austin I.S.D. - MIS Department
512.414.9715 (wk) / 512.935.5929 (pager)
From:
"Robertson Lee - lerobe" <Lee.Robertson@xxxxxxxxxx>
To:
<oracle-l@xxxxxxxxxxxxx>
Date:
11/17/2009 09:40 AM
Subject:
Stress my CPU's
Sent by:
oracle-l-bounce@xxxxxxxxxxxxx
Hi,
Oracle 10.2.0.3
AIX 6.1
I have been asked to do some stress testing and it is causing me stress ;-)
I need to really hammer the CPU's on the box as we want to test dynamically
allocating CPU's from another partition to handle the increased workload.
I am having difficulty getting the CPU's maxed out or indeed to even use more
than 2.
I have been running cartesian joins with various hints (parellism, using hashes
etc) but just cannot get the relevant amount of work generated. It tends to
spike at about 80% on a couple of CPU's then the result set comes back after
about 30 secs and CPU useage then obviously drops. This is a T & D server with
16 CPU's on it. Are there any clever tricks out there I can use to simulate a
full on workload??
TIA
Lee
Other related posts:
