Slackware 64 bit current.
8GB ram
Two (2) Western Digital Black edition 1 TB SATA III 7200 RPM drives
3.1.1 kernel
Phenom 3400MHz six core cpu

When I have a process that is maxing out the hard drives, such as a file copy process, the entire system becomes sluggish and responds poorly. If I have a 20GB file, such as a large database or a virtual machine disk image, and I want to copy it from one drive to another, the system becomes very sluggish and responds poorly during the file copy.

Are there ways to optimize this system such that the system can still be usable even when disk i/o is maxed out, and maybe to optimize disk speed?

Use the ionice command on the copy to perform the I/O at lower priority than foreground processes. For example:

The ionice command is for I/O as the nice command is for CPU. They can be combined as well:

That's not really a (long-term) viable solution.
I'd be surprised if Pat has pushed a 3.1.1 kernel out the door, so did you do it yourself ?. Might need some closer inspection of options selected. I have a couple of that level kernel lying around, I'll do some quick tests.
Maybe have a look at latencytop.

It does indeed depend on a lot of things, so if you could post your .config or try and see if it still happens with the generic kernel, it would help.

In many cases the I/O scheduler does have an impact on it, so changing that may help. You can change it dynamically by echoing the name of the scheduler into '/sys/block/sda/queue/scheduler'. Change 'sda' to whatever your disk is.

What filesystem are you using ?

Since you build your own kernel I would always recommend using a better cpu scheduler (BFS) and a better IO scheduler (BFQ). If that doesn't improve interactivity sufficiently you can help the system along by following macemoneta's advice. For instance, all my xterm sessions are automatically started in SCHED_IDLEPRIO and with ionice level 3. Whatever I do in an xterm session (using all virtual cores and all disks) has no noticeable effect on my desktop interactivity.

http://ck.kolivas.org/patches/3.0/3.1/3.1.0-ck2/
http://algo.ing.unimo.it/people/paol...es/3.1.0-v3r1/

Having said all that, if your copy process is a regular occurrence you might want to optimize the spindle layout by putting the system on a third disk or even a small SSD.

3.1.1 is on kernel.org right now, and it came out just this last week.

It is normal for the kernel to swap out processes to re-allocate the memory for disk caching. If you don't want the kernel to do this, lower the value in

/proc/sys/vm/swappiness

or set it to zero.
Ed

What specifically is your system doing when you say it's sluggish? Is the CPU pinned? Running out of memory and swapping? Are you trying to do something that is competing for the busy disk(s)? If you don't measure it you won't know what the problem really is.
-mark

I have one of these hard drives mentioned by the OP.

I only see slugishness when I use KDE to copy/paste large files. If I use midnight commander (my default file manager) I don't have this issue.

I'm talking in reference to slack 13.37 fwiw.