Phillip Bruce wrote:> . Keep in mind that swapfile uses the UFS > filesystem which is an overhead and if your using heavy databases please> use this as a temporary solution and not a long term solution. It can > be used to guage how much swap should be consider. Using raw> disk space has better performance than using swapfile.Can someone quantify the overhead of swap on UFS? I seem to recall the effect was marginal, on the order of 1% compared to raw disk. (in timing? throughput?) The overhead of swap on ZFS would be interesting as well. But you wouldn''t want to include the swap file in your checkpoints. Yet another application for the sticky bit? Cheers, Henk
Henk Langeveld wrote:> Phillip Bruce wrote: >> . Keep in mind that swapfile uses the UFS filesystem which is an >> overhead and if your using heavy databases please > > use this as a temporary solution and not a long term solution. It can > > be used to guage how much swap should be consider. Using raw >> disk space has better performance than using swapfile. > > > Can someone quantify the overhead of swap on UFS? I seem to recall > the effect was marginal, on the order of 1% compared to raw disk. > (in timing? throughput?) > > The overhead of swap on ZFS would be interesting as well. But you > wouldn''t want to include the swap file in your checkpoints. > Yet another application for the sticky bit?Why would you swap to a file on ZFS when you can so easily use a zvol? - Bart -- Bart Smaalders Solaris Kernel Performance barts at cyber.eng.sun.com http://blogs.sun.com/barts
On Mon, 2006-03-06 at 19:03 +0100, Henk Langeveld wrote:> Phillip Bruce wrote: > > . Keep in mind that swapfile uses the UFS > > filesystem which is an overhead and if your using heavy databases please > > use this as a temporary solution and not a long term solution. It can > > be used to guage how much swap should be consider. Using raw > > disk space has better performance than using swapfile. > > > Can someone quantify the overhead of swap on UFS? I seem to recall > the effect was marginal, on the order of 1% compared to raw disk. > (in timing? throughput?)If your swap device is in use, then your performance will suck. Period. End of discussion. -- richard
>> The overhead of swap on ZFS would be interesting as well. But you >> wouldn''t want to include the swap file in your checkpoints. >> Yet another application for the sticky bit?Bart Smaalders wrote:> Why would you swap to a file on ZFS when you can so easily > use a zvol?Right. zvol. Forgot about that.