server benchmarking util

topher · 9 Oct 2008 at 13:35

Guys, I am looking for a good but free benchmark util!

any recommendations? currenly using Nuclearus to compare 2 servers.

bigredshark · 9 Oct 2008 at 13:55

Personally I'd never benchmark servers in general performance, it's just pointless, benchmark them according to what app they're going to run...

topher · 9 Oct 2008 at 14:02

I agree but I am being pressured from above!

madman045 · 9 Oct 2008 at 14:06

Pressured to display what though? As bigredshark has said, you need to test the server for the role its going to perform, such as if it was going to be an exchange server, you can install exchange and then simulate SMTP and client usage to see how well it copes.

Benchmark software is designed for the desktop, but can be useful for things such as disk access issues if looking to upgrade.

bigredshark · 9 Oct 2008 at 15:05

madman045 said:
Pressured to display what though? As bigredshark has said, you need to test the server for the role its going to perform, such as if it was going to be an exchange server, you can install exchange and then simulate SMTP and client usage to see how well it copes.

Benchmark software is designed for the desktop, but can be useful for things such as disk access issues if looking to upgrade.

Jetstress is a good test, but it's only applicable to exchange really and it feeds on disk IO. There are some good tests for SQL as well.

You could run prime or something similar to show processor performance and a disk benchmark to show throughput and latency but it's all fairly meaningless without an app in mind.

morfmedia · 9 Oct 2008 at 19:36

SQL IO stress is useful for SQL testing, likewise IOMeter for disk I/O benchmarking.....

As above, you need to benchmark for intended use.

topher · 9 Oct 2008 at 20:15

cheers for the reply guys, gonna use SQLIOSim to and compare the results on 2 servers

3yr 2x dual core cpu
2month old 2x quad core cpu

the development guys are complaining the SQL queries are running slower on the new server compared to the old. however they moved the database's from sql 2000 to 2005, so perhaps its a SQL issue!

thanks again

DustyMiller · 9 Oct 2008 at 20:18

how much ram have both boxes got ???? SQL Loves Ram

bigredshark · 9 Oct 2008 at 21:06

SQL loves disk IO too, and SQL2005 is exceptionally fast in my experience, it's been a big boost over 2000 for us.

madman045 · 9 Oct 2008 at 22:53

Indeed, always best with SQL to have the logs and the db on different arrays where possible, usually the DB on a 15k SAS raid 10 array and the OS/Logs on a Raid 1 array

s0ck · 9 Oct 2008 at 23:33

What is the array config of both servers?

bigredshark · 10 Oct 2008 at 19:59

madman045 said:
Indeed, always best with SQL to have the logs and the db on different arrays where possible, usually the DB on a 15k SAS raid 10 array and the OS/Logs on a Raid 1 array

It's a good point in really high load systems, I've dumped both on a 6 disk RAID10 (72GB 15k SAS drives) in the past though and it was quick. Still, as you say it is best practice but unless it's really high load it probably isn't the performance bottleneck...

akakjs · 11 Oct 2008 at 13:17

topher said:
the development guys are complaining the SQL queries are running slower on the new server compared to the old. however they moved the database's from sql 2000 to 2005, so perhaps its a SQL issue!

Have a look at The SQL Performance Dashboard it should let you identify the queries with issues and see if the problems are I/O or CPU bound. It could be as easy as a missing index when you changed servers.

If you get the wait statistics for same application on the old and new server, you should be able to prove your point. Real world information beats benchmarks everytime.

[edit]ok now I'm kicking myself.... I forgot SQL 2000 can't run the performance dashboard. however it should still help you debug your performance problems.[/edit]

akakjs