I'm playing around with Intel's TBB (parallel run-time), and I'm trying to nail down the source of a substantial slow-down related to disk I/O. Specifically, the more threads I have trying to read from a pool of files, the worse my performance gets. I want to track down a way of seeing exactly what the file system is doing in response. Ideally, I'd get something like a log of (timestamp, sector) reads, to see if I'm getting murdered on seeks, or something. Thoughts?