On Mon, Aug 20, 2007 at 05:01:51PM +0200, Roch - PAE wrote: > > Peter C. Norton writes: > > On Wed, Aug 15, 2007 at 04:57:05PM -0700, adrian cockcroft wrote: > > > How fast do disks turn? You get one page per revolution. Adding more swap > > > disks would only help if there was more than one thread trying to read > the > > > data. Ultra 1 had a nice fast 7200rpm SCSI disk... > > > Adrian > > > > In an x4100, they turn at 10,000rpm. While the OS may only get one page > > per request, it's not impossible to get a sequentially written file > > with sequential blocks in the same track, and I'd be surprised if > > on-disk read-ahead cache didn't get invoked at least somewhat > > mitigating this. > > > > As Adrian said. A /tmp file will be spread all over a swap > disk in 8K chunks because of the way it was swapped out. We > thus expect to make progress at 8K per rotation per swap > disk. Or a very rough MB/s per disk. > > The feeling that swapin did not progress reflects the fact > that disk rotation progress at a much slower pace than the > rest. > > BTW, mount_tmpfs(1M) has a "-o size=" option. > A good best practice to have in order to prevent swapping > due to excessive /tmp usage. > > -r
I appreciate the recommendataion that swap use be minimized, and I agree that without swap, the VM system is not going to have any problems with swap. However, if swap is included and considered a normal part of the system, then its performance, while less than the optimal situation of having enough RAM, shouldn't be as poor as the apparently hard limit of 2.5 mb/sec peak when swapping more than about 100mb back in. There are innumerable cases where it is legitimate to expect a process to swap out, but be able to page in chunks of memory in a timely fashion. Since memory in systems has grown so large, the VM shouldn't consider 2.5mb/sec enough to accomodate a realistic workload. On most server disks, this kind of performance can't even come from a pathologically bad case (assuming that every requested block is only visited once). Modern hardware has caching layers on the disk, in the controller, and request queueing. However, I now notice that the only kernel tunables relate to the speed at which data is written to VM - there is nothing relating to swap-out speed. I have a feeling that someone put in a governer on that a long ime ago, and it's never become a tunable. Perhaps this isn't the right list for this, since swap is often seen as the enemy of performance. Is there a more appopriate list/group? Thanks, -Peter -- The 5 year plan: In five years we'll make up another plan. Or just re-use this one. _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
