06-19-2012 08:56 PM
Dears ,
i get the AVG Page fault/sec = 1500 , is it acceptable or high value ?
06-19-2012 10:21 PM
That metric is unimportant.
Monitor the following instead
network bandwidth (including FC HBA's too)
CPU
RAM utilization.
disk queue length
For the first 3, use the 70% rule. Never let those get over 70% for sustained periods during heavy activiy.
Disk queue length depends on the drive counts, but it shouldn't go over 2. But this is based on the average, which is over divided by the disk count.
So if RAID, you have to do queue length / spindle count = hopefully <2
06-19-2012 10:55 PM
Hello,
Windows performance counter “Page fault/sec” includes both soft and hard page faults. Soft page faults are not much of a concern as the pages are already available in RAM, just that they are not part of the processes accessing those pages. It is the hard page fault that impacts performance because in this case data is read from the page file on the hard-disk. Hard-disk seeks and reads are slow. It will be useful to know the below things to answer the original question.
*Avarage “Page Reads/sec”, Avg “Pages Input/sec” and avg “Page fault/sec”. Ideally they should be averaged over the time when BE jobs are run.
* Physical memory size of the server.
Thanks.