A learned a long time ago about the vmstat 1 command, which will allow you to watch some of the action that’s happening in your system. It’s a useful command, and I find myself relying on it most often when I want to get an idea of drive input and output.
But for an application that is dedicated to watching drive throughput as it occurs, by process, iotop is a better solution.
In the vein of iftop, htop and perhaps even iptraf, iotop looks a lot like the age-old top, and behaves similarly too. Process by process, you get a full list of everything running and what it requires in disk access. Aside from that there are some filters available, a batch mode that will run uninterrupted, sorting options and so forth, as you might expect.
It’s a python program, and as mentioned in the past, programs that run through python or other interpreters can be more taxing on underdog hardware. I get the feeling this one is fairly lightweight though, and considering it might help you pinpoint that annoying drive hog somewhere, it’s probably worth the resources it requires.
P.S.: fbterm in the background once more. Suddenly I have a use for all my old wallpaper again. …
Motho ke motho ka botho 
question: do you know of any way to monitor disk I/O on a cluster that uses a shared file-system? these tools (top, vmstat) appear to only show you disk I/O by the computer you’re currently using… but if that filesystem is shared by a number (or hundreds) of other computers, do you know a good way to watch I/O in real-time?
I am afraid I’m out of my depth when it comes to clustered or shared file systems. Perhaps someone else can offer a suggestion. … 😦
Iotop is another reason not to use Centos.
The latest version will not work with Centos 5.x as the kernel is too old and is the version of python (yet another useful tool you will miss if you choose centos.)