On Wed, Aug 11, 2004 at 02:23:24AM -0500, Phil Dybvig wrote: > > >1 problem) my SATA drive is too slow now (14MB/s - by hdparm) > > new 2.6.7 kernel uses SCSI emulation [/dev/sda] > > previous 2.6.6 kernel (50 MB/s) uses standard way [/dev/hda] > > I just did a quick test copying with cp a 550.8 meg file (actual > megs not decimal megs) from one SATA drive to another and the time > command. I did this three times each in the old kernel and the new > kernel, with the following results (times in seconds): > > 2.6.6-1.435.2.3 2.6.7-1.494.2.2 > real user sys real user sys > 6.165 0.013 1.649 12.016 0.025 1.901 > 6.892 0.024 1.753 13.649 0.022 2.001 > 13.221 0.026 1.667 5.155 0.022 1.851 Please check the drive specifications in detail. I was looking at the specs of a disk drive recently and noted that the media bit rate could differ by a factor of two depending on the region of the disk. The media bit rate was much slower in both cases than the system interface. Caching of data in memory can further confuse things. Looking at your table I see a 2:1 ratio and then a flip for the speed winner of from ..435 to ..494. If the factor of two I noticed in the documentation I looked at exists on your drive... the difference might be explained by zone locality differences of the data blocks on the disk. If I take a risk and take all 2:1 differences as a locality difference then I might infer that ...494 might be faster by a percent except that the rate was not exactly 2:1. One way to test this might be to partition the disk into ten equal regions and test a files system each. Or to do raw IO to specific regions without the overhead of a files system. Since modern disks are complex and smart, testing can be difficult. SATA disks will be 'smart'.... -- T o m M i t c h e l l Just say no to 74LS73 in 2004
