On Jan 8, 2004, at 11:31 AM, Jan E. Schotsman wrote: > I find it hard to believe there is no relation between sector numbers > reported by a controller and actual position on disk. Surely it is > important for disk performance that a controller tries to realize a > request > for contiguous disk space as a set of sectors that is close to each > other > in some sense? Please explain, Dmytro. 1. multi-platter disks 2. read/write coalescing inside the controller 3. flaw mapping inside the controller (it is much cheaper to make N% bigger disks disks with N% flaws, than to make flawless disks - same technique is used inside memory and cpu chips too). 4. spiral writes Please refer to the disk tutorial url that I provided a few days ago, if these terms are new to you. Bottom line: the disk controller believes it can do a MUCH better job of laying out the disk than the kernel can, and it is usually right. Performance competition is intense in the HDD world, and the manufacturers take this edge because they know they can make their controller do a better job than to just blindly follow write instructions from the device driver. Peter