[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