Re: help:vmstat show constently po about 400

Hi Joe,

vmstat "page-out" operations are normal, as the vm prepares for the possibility of a true page fault, when the page contacts are re-read into RAM.

I generally watch the vmstat interaction between sr (scan rate) (Solaris only) pi (Page in) and po (page out).

The goal is to detect if your server is "truly" swapping RAM regions, and it is not easy to see!

First, what UNIX are you on? If you are on Solaris, you can correlate the vmstat sr (scan rate) to detect "true" page-ins.

As for SGA size, I compute the HWN of connected user RAM as follows:

TOTAL PGA RAM = #users * 2meg stack * sort_area_size * hash_area_size

For example, if your SGA is 800 meg and you have a PGA size of 10 meg each, then you will start swapping when total RAM exceeds about 1800 meg (we must reserve 20% RAM for the kernel).

In this example, you will probably start paging at about 100 connected users.

Hope this helps . . . . Received on Sat Jun 29 2002 - 08:36:47 CDT