CPU/Mobo Hard-core database work

L

lide

Disciple
I'm looking for a configuration for a workstation that will be used for intensive database work. I am working largely with mysql on databases with well in excess of 15 million rows. Any simple query with links between tables takes in excess of 15 minutes. A 'Group by' query can take almost an hour

My current config is an HP with:
CPU: Intel E6550
RAM: 1GB
HDD: 7200 RPM

I have a few questions:
1. How can I identify the bottleneck in the current system so that I can check if simply upgrading some components will help?
2. If I was to build a new system what config would be best? Remember I don't need any graphics just pure database performance.
3. Why does Dell use a Xeon for their workstations (link provided below) when I thought Xeons are used for servers? How does a Xeon differ from our regular core2 and quad core processors?

Dell Precision T7500 Tower Workstation Desktop Computer Product Details
 
Xeons are server class cpus and lower end ones are used in workstations also. You can assemble your own server buy buying server class boards like Intel Server Board S5000VSA - Overview
or
Intel® Xeon® Processor 3000 Series, Features
boards with Xeon or other compatible CPUs.

Check their Cache L2, L1, etc.

Regarding Database, you didn't which database you are using. Anyway, you can 'profile' databases according to your use /purpose. For the no. of records you are processing, your RAM is very low. At least I know that MS SQL support profiling.
 
Prolly the HDD coz mycql is not that CPU intensive. Add loads of ram and load the entire DB into your RAM. That should give you lightning fast results.
 
I overlooked that you mentioned mysql. Pls. try to shift to postgresql. It will help a lot in performance especially when you have large no. of records.

Its well known that mysql is not upto the mark in performance in comparison with postgresql when large no. records are involved. Test your database thoroughly in postgresql, with increased amout of RAM obviously.
 
Thanks guys. More RAM it is then.

I had tried postgre recently but didn't make any difference so switched back. I guess that's because the RAM is really the bottleneck rather than the application. Plus there was the hassle of rewriting my queries because of the somewhat different SQL.
 
Back
Top