disk io performance EVA8000 with blade860c hpux 11.31

[utku]

hi all;
i have a BL860c itanium 4 core 8GB RAM blade server (hpux 11.31) which is connected to an EVA8000 box. EVA has 240*300GB 10Krpm disks and 3 disk groups (3*80disks). 3*2TB disks have been presented to that blade and a 5TB file system has been created for oracle (striped). this server is used as a standby oracle db of a production one (oracle 9.2.0.6.0). archive logs are transferred to that 5TB file system from prod and applied under same place. during IO operation to that file system, about %25 io wait occures despite sar disk and evaperf vdg outputs are considerable. dba guys also complain about the slowness of the server. ı tried scsimgr queu depth, load balancing options, "ioerror=mwdisable,mincache=direct,delaylog,convosync=direct" mount options but no changes. sar disk, cpu and eva perf vdg outputs during same 5 minutes with 1 second interval are in the attachment. any comment appreciated.

[utku]

sardisk output

[utku]

sarcpu output

[Vivek_Pendse · Accepted solution]

Hi,

With your sar o/p, the avg wait time is always less or approx "0" & the avg serv time is always going high for some disks.
Avg serv time, where you have to do something outside hp-ux box.
Will recommend you to check/compare the sar o/p with EVA perf for some disks where av serv time exceeding.
May be, either any policy defined for those disks, then need to change that. Also, check the wait IO in that time.
Also, keep in mind that the EVA is Async A-A.

Hope this helps.

Thanks,
Vivek