> You're right about the rotational speed of the disk, but > don't the sectors get wider as you get out there? If you > move the head, the worst case is that you have to wait for a > full revolution once you hit the right track, but angular > velocity is the same everywhere, so I guess I'm not seeing > how this gives you faster seek times. They do, but it does have an effect. Maybe because of the increased throughput and the drive can figure out faster if it's found what it's looking for? In one of my CS classes at the U, we spent a couple of weeks on disks and this was one of the things we covered. I don't remember which class it was though. > > > I typically have a ton of mail > > on my postfix boxes, so moving my spool directory to the last > > partition on the disk has increased my performance by quite a bit. > > I've also started doing remote logging from all of my > mailservers. By > > doing this, postfix doesn't have to fight at all with > syslog for disk > > access, and that has also increased my performance. > > > > Basically, I went from an average of 160ms to inject a message into > > the queue using the default settings, to now about an > average of 35ms > > for each message injected after tweaking everything. > Installing the > > 20011127 snapshot brought me down from 60ms to 35ms just by itself. > > Doesn't this say that the disk partitioning had no measurable > effect? Of course, putting it on a device that isn't being > asked to fly off to other parts of the disk will be a big > speed up, but I thought you had that already. I did the disk thing before upgrading, and it dropped me from about 80ms to around 65ms for message injection to the queue. > I'm not arguing with you, just trying to get clear. Thanks, Phil Well, you can try testing it for yourself. Grab a copy of bonnie and run it on an unloaded system. Try a run on your first or second partition, and then try a run on your last partition (make sure it starts closer to the end of the disk than the beginning, or your results will be similar). Bonnie will give you number of seeks per second, you should run it with a filesize which is at least double what you have for memory and run 3 times to make sure you get the same results each time through. I didn't run bonnie to test it, I just have graphs of our email injection time over the last couple of weeks which decreased when I made the change. Jay > > -- > "Trying to do something with your life is like > sitting down to eat a moose." --Douglas Wood > _______________________________________________ > Twin Cities Linux Users Group Mailing List - Minneapolis/St. > Paul, Minnesota http://www.mn-linux.org > tclug-list at mn-linux.org > https://mailman.mn-> linux.org/mailman/listinfo/tclug-list >