I know there's a ton of different tools you can use to improve disk performance, power management etc. I'm just wondering how much tuning can you do until your just wasting time? After all, the new Linux Kernel has been released, with a bunch of new performance fixes.

I'm using a HP workstation. It seems to get a little warm.

So far I have:

dhparm(not even sure if the Kernel is configured to it)
readahead(seems to work fine)


What else is a must for Linux users?

Isn't that a balance between what the hardware can handle, what the (measured) increase in performance will be, what the implementation cost is and what you actually need it for? I mean if you need near-realtime response you accept the fact your hotrod comes with custom tyres (RT patches, scheduler) and without fenders (firewall, SE Linux, caching), car stereo or a door (DE/WM). On the other side of the spectrum, if yours is say a '82 Regency Brougham, then it prolly makes sense to try and wring every ounce of juice from the setup no matter the cost. If yours is a recent all-purpose one then I don't think you need to tweak a lot (Linux is quite efficient) unless you *know* you need it (patches, scheduler, VM, network, disk, caching).


It may be an 8-way SMP box, DDR2 is less energy-efficient compared to DDR3 (AFAIK), your case may be dusty or have bad airflow around components or just not enough active cooling. Saying it's getting warm w/o reference doesn't mean that much.