hp - oracle (dataware house system) has high i/o wait time

IO wait by itself is not a good metric to look at. A better metric is response time or service time. If the response time is reasonable for your workload, then I wouldn't worry about IO wait time at all.

IF %hit ratio is below 70% then in most of the cases it would indicate not a good performance at all but the best use of hit ratio is still to compare over a time to help alert you to a substatial change to a system on a given day.while there are some that have deprecated hit ratio.

and most important Hit ratio is not always accurate see the below mention scenario.
If u r utilizing hit ratios to measure performance,also note that during peak times the number of disk read is larger than the number of in-memory reads and thus the negative hit ratio is being computed in terms of deltas between read and logical reads.


Sanjeev Tanwar

You cannot get an accurate answer by measuring %hit only. But to measure the hit rate, you have look at the %hit and %write. Since all writes go to cache on Symmetrix, you can measure the read hit rate by getting the value between %hit and %write. And also having the read hit value low does not mean that there is a problem. Most OLTP applications IOs are more random and the hit rate is not a problem here. By increasing the cache, you increase the possibility of data being in the cache in the next attempt. But in your case if that is a DWH you have get more read hit rates. Depends according to you storage config(protection level, striped or concatenated metas), host config(lvm and filesystem config and parameters) and oracle(ASM or filesystem).
Also for HP-UX systems, wait io calculation is somehow different from AIX and Solaris. If there is only io bound processes in an HP-UX environment, it is so normal to get those high wait io values. So you have to look at the average queue and service time values for HP-UX systems.