Musing on memory useage

With all this talk of performance and memory recently I decided to take a look at the memory being used on my 800mhz 384Mb machine.

Just a snapshot taken today shows:

70,800K Memory
72,552K Peek Memory
91,452K Virtual Memory
25,680,562 page faults now 25,721,329 since I started typing tihs entry in the forum.

Now what puzzles me is the page faults because as I watch this figure increases quite dramtically in bursts adn I think this is what may be causing some of my performance problems. Also there is a thread about Handles and mine is sitting at 8258 and climbing. WD has been running 10.27c since 10:40:59 11/07/2005, ie just over 5 days.

This PC does run MySQL as well but it is not hitting memory nearly like this. It also hardly ever page faults although it has a larger VM but a smaller memory asize compared to WD.

Now in my mainframe days page faulting like this would be frowned upon from a performance aspect. Right now I cant increase memory in the m/c without buying some so I dont want to do that just yet until I am convinced there is nothing else needs doing. With WD page faulting like this and nothing else significant running apart from MySQL and nothing else page faulting I am wonmdering what WD is doing to get all these.

Anyone have any views? I’d be interested in thse figures from other people running slower m/cs as well.

Stuart

Here’s mine on a 733 MHz 256 meg P3 with W2K and WD 10.26t Main things running are WD, norton reatime virus protection, zone alarm, VNC server (and hypersnap to grab the screen). At the time of the snap WD was using 30,696 K. Where can I see page faults?

its those increasing handles that is the problem
maybe try turning functions off in wd until they stop increasing?
it could be some file error…
but not everyone is having that issue
its not a problem for me

You can add other fields to the display by doing View - Select Columns on the Processes tab.

Stuart

Interestingly the Handles have dropped back slightly and the only thing which has changed is the solar sensor is now reading 0. I’ve not done anything in WD at all.

Stuart

Thanks

ahh, i have added alot more info now too
i see my peak memory use is 107m
i have 120,000 page faults ( been going 1 week)
i see zone alarm has 14,000
what causes those??

I think it just means the program didn’t find the info it needed in RAM memory so it had to go get it from the info that had been moved to the virtual memory on disk.

My fault about handles, I was looking at the total count. The handles for WD is only about 191, MySQL is sitting at 4434. Still the system wide toatl is over 8000.

I dont know what is causing the page faults, other than the page needed to process is on disk and not in memory so you have to page it in before you can process the data or execute the code. I am not very familiar with Windows memory management, but logic says if you are page faulting a lot it can be either a code bug, an operating system problem or a lack of actual physical memory. What is needed to know about the code design is what actual pages are being faulted, it could be that a slight change in design could ensure that a frequently paged page could be made memory resident. If the pages being faulted are random then it is more likely to be a lack of physical resources which cause the problem.

Stuart

I had WD crash this morning. I just got home and re-started it 10 min ago and my page faults are at 267,000 in only ten min. Version 10.27C on old P II 450 mhz 192 Meg Ram Windows 2000.

Frank

I really don’t think page faults are anything to get excited about, that’s just the way that windows works.

How many handles do you have?

i would say the less physical memory, the more page faults

Handles are in the 5,800 range.

Frank

i only have 470 handles

Is that total, or just for weather display?

Using WD 10.26x here only have 179 handles for WD, using WM918 and 3,

…or just turn off indexing altogether and save some disk thrashing.

If you get a lot of page faults then that’s a sure sign that your PC is saying “Give me more RAM”.

Chris while I can agree with that in general throwing more RAM at a machine will not always fix the problem.

I have also found that what I thiought shold be a small program wdwatch has over 339 million page faults in the same time frame as WD itself. Now I am sure that as well as adding more RAM something could be done about such a small program causing this many page faults. Also every page fault is at least 1 I/O if not 2 or 3 depending on how good the page tracking on disk is.

Stuart

wdwatch is just checking a entry in the wdisplay.ini file…
now, i leave that ini file in memory, instead of crfeating/freeing it each time…
but doing that, every read of that ini file must be causing that page fault
but i dont think that is a problem