Hi,

Benchmarking is more of an art than a science at the end of the day. Most of 
the benchmarks out there like SPEC.org's CPU or LINPACK are really just a 
series of complex math problems that target integer and floating-point 
performance. They don't reflect real-world applications, and if anything only 
reflect "calculator envy". As such, I would say that the closest "real-world" 
benchmarks will actually run an application and simulate a workload. So 
depending on the application the machine is running you should target 
benchmarks that can really test the load and scaling for that app. Here are 
some links that might be of interest to folks out there:

Oracle: http://www.dba-oracle.com/t_benchmark_testing.htm
MySQL: http://dev.mysql.com/doc/refman/6.0/en/mysql-benchmarks.html
             http://sysbench.sourceforge.net/
Apache: http://httpd.apache.org/docs/2.0/programs/ab.html

Now many of the big benchmarks are unfortunately not free or unavailable to 
non-"partners". But if you have the money, I would recommend the following 
SPEC.org benchmarks:

JBB2005
jAppServer2004
WEB2009

Those benchmarks simulate real web and java workloads that you can run against 
your JVM, app server, and web server. 

Another route to go is to capture all the traffic to your production 
application, and replay it against your target benchmark server. Compare the 
performance and response time. Then double the traffic to see how well it 
handles the increased load. There are some tools such as Mercury that can do 
this. Ultimately, I would say this is really the best route for determining if 
a new server will perform better for your application than another.


 *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
Octave J. Orgeron
Solaris Virtualization Architect and Consultant
Web: http://unixconsole.blogspot.com
E-Mail: [email protected]
*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*



----- Original Message ----
From: Ben Rockwood <[email protected]>
To: [email protected]
Sent: Monday, June 29, 2009 1:41:29 AM
Subject: [sysadmin-discuss] CPU/Memory Benchmarking

Anyone have recommendations for solid CPU and Memory (separately)
benchmarking tools?  I've got several new systems I'm testing and I want
a solid tool for doing comparative analysis on the systems and am too
lazy to work up a series of real-world benchmarks.  I've yet to see
anything really good for Solaris.  Most everything out there is ancient.

Please note, I'm not looking for a SunVTS type stress test, I want
numbers for comparing system A against similarly configured system B. 
If it can go the extra mile and give me concrete memory
bandwidth/latency numbers or report CPU capability in *flops, all the
better.

benr.
_______________________________________________
sysadmin-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss



      
_______________________________________________
sysadmin-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss

Reply via email to