Page Faults

[Open Systems]

Hi Guys,

Hope someone can help, I'm trying to understand this glance screen for a poor performing server I'm looking at.

B3692A GlancePlus C.03.58.00 11:42:33 eagle 9000/800 Current Avg High
------------------------------------------------------------------------------------------------------------------
CPU Util S SRU U | 82% 76% 100%
Disk Util F F | 3% 9% 71%
Mem Util S SU UB B | 60% 60% 60%
Swap Util U UR R | 30% 30% 30%
------------------------------------------------------------------------------------------------------------------
MEMORY REPORT Users= 2
Event Current Cumulative Current Rate Cum Rate High Rate
--------------------------------------------------------------------------------
Page Faults 0 907635 0.0 496.4 5898.6
Page In 0 259599 0.0 142.0 1523.4
Page Out 0 0 0.0 0.0 0.0
KB Paged In 0kb 136kb 0.0 0.0 57.1
KB Paged Out 0kb 0kb 0.0 0.0 0.0
Reactivations 0 0 0.0 0.0 0.0
Deactivations 0 0 0.0 0.0 0.0
KB Deactivated 0kb 0kb 0.0 0.0 0.0
VM Reads 0 10 0.0 0.0 0.5
VM Writes 0 0 0.0 0.0 0.0

Total VM : 368.4mb Sys Mem : 211.0mb User Mem: 1.71gb Phys Mem: 4.00gb
Active VM: 164.9mb Buf Cache: 491.5mb Free Mem: 1.60gb


The basic understanding that I seem to missing is how can I be getting so many page faults / high paging in rate, when the server never seems to page anything out ?

Can anyone explain ?

Thanks

[Ralf Seefeldt]

Hi,

for me, it looks like the server has been Using or Blocking much more memory in the past.
That process(es) may have terminated, freing 40% of memory while the server still ocupies 30% of his swapspace.

Now the server could page in, but there seems to be a promlem, which I would search for at harddisk or memory first.

How long did glance run when this snapshoot was made?

I have never seen such a high PageFault rate.

I hope this helps a little

Bye
Ralf

[Open Systems]

Hi,

That snapshot was made over around 30mins, I've been monitoring the server for around an hour now, and it's still paging in heavily, with no page outs.

Can a process page itself out, without using the vhand process ??

[Ralf Seefeldt]

Hi,

I don't think, a process can page out itself without using the vhand process - but I am not sure.

What is the uptime? Would it be possible to reboot the server and see, if this helps?
Are there any processes with high acumulated processortime or other unnormal habits?

Bye

[Open Systems]

Think I may have worked this out - got a document that describes the Page Faults parameter as also being incremented when a new process starts, and finds text pages resident in memory from a previous incarnation that it chooses to re-use. So the traditional understanding of page faults is incorrect - ie: it doesn't always relate to memory load.

Thanks for the help.

[Navin Bhat_2]

I dont see a problem here necessarily. The number of forks on the system can also choose the amount of pageins occuring. You are not experiencing any memory pressure and there are no page outs. How are you quantifying poor performance? Please post the output of "vmstat" also.

There are no process deactivations/reactivations either.

It is normal for a system to see soft and hard page faults.

Just because CPU util is high does not mean you have a performance problem.

[John Kittel]

In agreement with Mr. Bhat's response, but attempting to explain "between the lines" a bit for you, .... the system does not necessarily need to swap a page out just in order to be able to read a page in. Many pages are read in, used, then the process is done, the memory goes back to free page pool ( whatever correct technical term for this is), and there is always plenty of free memory for new pages to be read into.

- John

[John Kittel]

Also, you may want to have a look at HP's Memory Management white paper; it discusses this sort of thing.

http://docs.hp.com/cgi-bin/otsearch/getfile?id=/hpux/onlinedocs/os/11i/mem_mgt.html&searchterms=management%7cwhite%7cpaper%7cmemory&queryid=20040603-172912