[CentOS] Adding RAM
Matt
lm7812 at gmail.com
Thu Dec 18 15:37:20 UTC 2008
>> 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
> Like this:
>
> kernel /vmlinuz-2.6.9-78.0.8.ELsmp ro root=/dev/VolGroup00/LogVol00 elevator=deadline
>
> And this change will be for System Wide.
Doing this among other things helped quite a lot. My iostat -x %util
dropped from ~ 60% to ~ 22% now. Of course at same time I updated
from dual-core 2Ghz CPU to quad-core 2.4Ghz. Also swapped motherboard
from one with an "Intel ICH9R" Southbridge to a "Intel ICH7R" since I
heard CentOS 4.x did not have drivers in kernel for ICH9R.
Matt
More information about the CentOS
mailing list