On Sun, 2004-09-12 at 23:44, Mark Knecht wrote: > On Sun, 2004-09-12 at 19:49, Cournapeau David (ENST) wrote: > > Mark Knecht wrote: > > > > > While none of this directly applicable to what you are experiencing, > > >when I read your note it seemed similar enough to go ahead and write > > >back. It might not only be the type of files system, but also where on > > >the drive you are writing. It's not always that well known, but most > > >drives are 30-50% slower at the far end of the drive. Most drives are > > >speed rated when they are empty. > > > > > > > > Without knowing anything about the details of > > hard drive conception, I would think that ALL > > drives are much slower for tracks near the > > axe than on the border of the disks, since > > the speed is proportional to the distance track->axe. > > This is inherent to any system using a disk and having > > constant angular speed. > > > > I know that even windows > > 95 tried to put the data on the borders of the disks of > > an hard drive as long as it could. I am pretty sure linux does the same. > > > Yep - that, and also that as a drive becomes more full it also naturally > becomes more fragmented, and fragmentation causes more drive head > movement seeking out places to put data. > Many file systems also switch from their normal, fast layout algorithms to a slower but more space efficient scheme when they reach 90% full. I believe the BSD FFS was the first file system to do this, and large parts of Windows are known to be copied from BSD. Lee
