looking to induce high i/o read/wait for testing
 

I have some systems which are seeing a high I/O wait (in iostat -x, very high rsec/s). I have some parameters in mind for improving this, but currently I can't modify the systems directly. I was wondering if anyone could recommend a method (perhaps some sort of benchmark tool) which would artificially induce high I/O wait on a spare system so that I could try tinkering.

I'm using CentOS 6.0 in a vmware/ESX environment, if that helps.

Thanks!

In my (limited...) experience, tests of this type are usually done using special testing libraries which (programmatically) impose random wait-times. In other words, they do the I/O and then dawdle around for some random but programmable amount of time before returning. This allows you to simulate the effects upon a program of "murderously bad I/O thrashing" without actually incurring it, which of course "in real life" would make the whole damm system so unstable as to be unusable. (And you do it all, all very repeatably, in user-mode.)