And that's a good point.

You should, at a minimum, have sysstat (the Linux sar package)
installed on all of your servers and monitor that. Also, do some
remote SNMP performance monitoring to start seeing where your
bottlenecks are. Keep in mind that IO issues may appear as CPU issues
at times.

For a database server, the most important element is memory. I would
rather a single CPU database server with 8 GB of RAM than a quad-CPU
system with 4 GB of RAM.

I'm not sure where we are at on multicore vs. SMP when it comes
to databases. That would be interesting research for someone here.

The O'Reilly book on performance tuning is still my favorite. I have a
Linux-specific text somewhere that I keep meaning to read, but that
requires a bit more time..

Anyway, if you don't have numbers when tuning for "performance", you
are going to just be guessing. :-)

Oh, and finally, do realize that there is a difference between local
performance counters and end-to-end performance testing. An important
baseline is to see the min, max, and average for each SQL op (using a
large testset) while your server is at rest, maxed out, and operating
normally with clients. Local performance counters should help you tune
your system based on your end-to-end testing. The local counters have
no real significance otherwise. I could care less if my machine is
tagging the CPU at times if my client applications don't notice a
performance issue, but if you cared only about the local counters, you
may spend your time working on that needlessly.

---
Puryear Information Technology, LLC
Baton Rouge, LA * 225-706-8414
http://www.puryear-it.com

Author:
  "Best Practices for Managing Linux and UNIX Servers"
  "Spam Fighting and Email Security in the 21st Century"

Download your free copies:
  http://www.puryear-it.com/publications.htm


Sunday, January 14, 2007, 10:58:58 PM, you wrote:

> What's it for?  Why do you think you need more than one machine to meet the
> load?

> On Sunday 14 January 2007 20:24, Brad Bendily wrote:
>> cluster mysql for high-availability and possibly load-balancing.

> _______________________________________________
> General mailing list
> General at brlug.net
> http://mail.brlug.net/mailman/listinfo/general_brlug.net


Reply via email to