[cdelgadop]

Hi

I'm having some performance problems with one LPAR running AIX 5.3 ML02 with Oracle DBs (up to 4 on the same server). I use nmon in order to monitor my AIX systems and i'm spotting very high numbers for the paging to file system (with almost zero to paging space). I've seen some high mumbers for the lrud program which is consuming up to 10% of CPU.

I've run a filemon command for 10 mins and found that the most active files are related to one .dbf file. Best solution is to analize pattern of I/O and optimize by adding another disk to the LV and strip the I/O. But i've also noted high I/O to the /unix file (this is the kernel). Why this behavior ??, in which situations is the kernel read so such high I/O could happen.

Thanks in advanced

[cdn_borne]

10% is not that high. Is this a Transactional DB or reporting???

[cdelgadop]

Hi

Thanks for your answer. There're many DB on that LPAR, different in behavior. Some are OLTP and some DSS. They're on different LVM structures (but on the same HBAs).

I think i can improve the I/O by mounting BD FS as cio devices in order to emulate raw devices. This config should decrease the lrud activity, what do you think ??

[cdn_borne]

CIO will help with writing. You might also take a look at your jfs2 log slice. If it is on the same pv/vg that your seeing high paging, it will become a hot spot.

Regards