Re: excessive swap-in time
- From: phil-news-nospam@xxxxxxxx
- Date: 17 Mar 2008 13:07:41 GMT
On Mon, 17 Mar 2008 10:19:30 +0100 Rainer Weikusat <rweikusat@xxxxxxxxxxx> wrote:
| It is actually 'pretty darn easy': If there is ever signficant paging
| activity, the system had too little physical RAM at this point in
| time. If there is always significant paging activity, the system has
| simply too little RAM for its usual workload. The solution is (to
| repeat this again), to either add RAM until the regular paging
| activity stops or to reduce the virtual memory usage. There is no
| third option, except if counting 'complaining about problems one does
| not want to fix' as an option.
Ultimately, this is true. But the timing numbers suggest that the kernel
is not distinguishing well between active and inactive processes. What
it should do is be sure to NOT steal from the top active processes based
on how many of them can fit in RAM. With processes ordered from most
active to least active, add up their average footprint (not the whole VM
size), plus the typical buffer size needed, until the RAM size is reached.
The processes up to this point are then flagged as "most active" and any
page stealing should avoid them.
It would be nice if there was a kernel function, or even just an uncatchable
kill signal, that means "swap this process out" or "swap this process in".
Then I could force out all the other processes manually and see what happens.
--
|---------------------------------------/----------------------------------|
| Phil Howard KA9WGN (ka9wgn.ham.org) / Do not send to the address below |
| first name lower case at ipal.net / spamtrap-2008-03-17-0801@xxxxxxxx |
|------------------------------------/-------------------------------------|
.
- References:
- excessive swap-in time
- From: phil-news-nospam
- Re: excessive swap-in time
- From: Pierre Asselin
- Re: excessive swap-in time
- From: phil-news-nospam
- Re: excessive swap-in time
- From: Pierre Asselin
- Re: excessive swap-in time
- From: Rainer Weikusat
- excessive swap-in time
- Prev by Date: Re: excessive swap-in time
- Next by Date: Re: excessive swap-in time
- Previous by thread: Re: excessive swap-in time
- Next by thread: Re: excessive swap-in time
- Index(es):
Relevant Pages
|