[lug] Suggestions for Burn-In testing

[George Sexton]

Thanks for the suggestions. I think they've put me on the right track.

The app is a Java Servlet application. I'm running 50 sites on a P3 500 with
512MB of RAM given over to the JVM. The new machine is a P4 3.0 GHz w/
2048MB of Dual Channel DDR 400 RAM, and a pair of Serial ATA drives using
software mirroring.

I think the approach I'll go with is a combination of disk stress testing,
and application testing using Jakarta Jmeter to exercise the application.
The disk testing will help ease my mind about using Kernel software RAID,
and the application testing should flush out any JVM issues/networking
issues.

George Sexton
MH Software, Inc.
http://www.mhsoftware.com/
Voice: 303 438 9585
  

> -----Original Message-----
> From: lug-bounces at lug.boulder.co.us 
> [mailto:lug-bounces at lug.boulder.co.us] On Behalf Of Michael J. Hammel
> Sent: Thursday, July 07, 2005 1:57 PM
> To: Boulder (Colorado) Linux Users Group -- General Mailing List
> Subject: Re: [lug] Suggestions for Burn-In testing
> 
> On Thu, 2005-07-07 at 12:48 -0600, George Sexton wrote:
> > I've run Memtest86 on it, but I was thinking about some 
> sort of software
> > that could do some disk write loading, and misc. stability 
> tests. I'd like
> > to be pretty confident the box is going to be stable before 
> I spend the time
> > doing the final configuration, and actually put it in production.
> > 
> > Does anyone have any suggestions on what I could use?
> 
> For general system stability, including CPU and Memory tests, try the
> UnixBench benchmark.  For network throughput you can try ttcp or
> netbench, though these probably aren't necessary for system stability.
> For disk I/O try bonnie or bonnie++.  
> 
> All of these are linked from the Linux Benchmarking web site:
> http://lbs.sourceforge.net/
> 
> > The box will be hosting our calendar software for 50 
> customers (initially),
> > and hopefully up to 200 before it maxes out.
> 
> Those numbers won't be a problem for TCP or the web server.  However,
> memory usage and swap space usage are determined by the application.
> Heavy PHP use might chew up memory under the web server, for example.
> There is a benchmark called WebStone that might help here but 
> I've never
> tried it.
> 
> If you've got multiple CPU's or cores you can try running multiple
> copies of UnixBench.  Each copy should chew up close to 100% of a CPU.
> I've run three copies on a 4 core processor (leaving one to 
> keep an eye
> on things with "top") and the kernel managed it just fine.
> 
> Don't know of any thread benchmarks, but if you run into any let me
> know.  I may need them later when working with MPI.
> -- 
> Michael J. Hammel           |
> The Graphics Muse           |           Assassins do it from behind.
> mjhammel at graphics-muse.org  |
> http://www.graphics-muse.com 
> 
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
> 
>