On Dec 8, 2008, at 4:55 PM, Matt <lm7812 at gmail.com> wrote: >>> Cpu0 : 5.3% us, 4.0% sy, 0.0% ni, 81.5% id, 9.3% wa, 0.0% >>> hi, 0.0% si >>> Cpu1 : 11.3% us, 8.3% sy, 0.0% ni, 1.7% id, 78.1% wa, 0.7% >>> hi, 0.0% si >>> Mem: 8309188k total, 4761352k used, 3547836k free, 451464k >>> buffers >>> Swap: 2031608k total, 192k used, 2031416k free, 1564316k >>> cached >> >> If this is typical, you don't need more RAM... >> cpu1 is waiting a lot, I/O bottleneck? > > A bit of bottle neck. > > Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s rkB/s wkB/s > avgrq-sz avgqu-sz await svctm %util > sda 0.38 176.63 70.32 78.26 813.46 2044.82 406.73 1022.41 > 19.24 0.40 19.17 4.04 60.07 > sda1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 5.28 0.00 23.61 19.33 0.00 > sda2 0.38 176.63 70.32 78.26 813.45 2044.82 406.73 1022.41 > 19.24 0.40 19.17 4.04 60.07 > dm-0 0.00 0.00 70.71 255.60 813.45 2044.82 406.73 > 1022.41 8.76 2.90 8.87 1.84 60.10 > dm-1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 8.00 0.00 64.20 11.38 0.00 Try setting the scheduler to 'deadline' and see if the queue sizes shrink. No raid1? Besides adding redundancy, it can help with read performance. I would probably put the mail on a raid 10 though if I had 4 disks to do so. -Ross