Tuning an overview
When tuning, remember that there is no “Magic” setting, no single set of tuning parameters that cover all situations. Each server will have it’s own bottlenecks. This can be caused by many things, here are a few examples.
- Software & Applications
- Hardware (cpu, memory, architecture, bus speed, etc.)
- LOAD end user, application, and system overhead.
- Outside factors (network, san, etc.)
When tuning a system there are several main items to keep in mind.
- Each system is unique
- Alter 1 thing at a time
- Use the settings recommended by your software / hardware provider
- Alter one setting at a time. This will give you the opportunity to see if that change makes an impact. Making to many changes may degrade performance.
- Test after each change. Be sure to gather enough performance data so you can determine if the change had a positive or negative impact.
- Did I say test?? I can’t stress this enough.
- Document, Document, Document. Document every change you make. This will save you time and sanity if you need to back out some changes, and allow for the change to be easily duplicated.
- Shutdown unneeded services. Take a look at the output of “chkconfig -l | grep on” I bet you there are a few things we can shutoff.
- Gnome ???? on a web server?? init 3 anyone??
- iptables, SuSe-Firewall, RedHat Firewall, etc. If you don’t have unneeded services running why do you need a firewall to restrict access to them???
- SELinux, AppArmor?? are big performance killers and usally not that important.
