1 |
My previous server OS choice was Red Hat, which did often change to swap. I
|
2 |
though Gentoo did it too.
|
3 |
|
4 |
Thanks,
|
5 |
|
6 |
Tom
|
7 |
|
8 |
On Fri Jul 30 13:22 , 'Florian Koenig' <k0255220@××××××××××××××××××××.at> sent:
|
9 |
|
10 |
>> > Also, when I do |
11 |
>> > |
12 |
>> > free |
13 |
>> > |
14 |
>> > it shows no swap is used even though physical memory level is 98%. |
15 |
>> |
16 |
>> free shows no swap used on my system either. and 90% or more physical |
17 |
>> memory use is standard for the linux kernel. |
18 |
> |
19 |
>Exactly. In contrast to Windows NT/2000/XP, which tries to swap out as much |
20 |
pages as possible, Linux tries to keep the memory mostly filled. Lots of the
|
21 |
memory gets used as cache and buffer so you might actually gain some performance.
|
22 |
> |
23 |
>The difference between Windows and Linux stems from their different origins. |
24 |
Linux/Unix was always meant to be a server system where the set of running
|
25 |
processes doesn't change that often. Therefore it tries to keep pages in memory
|
26 |
for as long as possible.
|
27 |
> |
28 |
>Windows' approach to the problem is geared to desktop performance and so it |
29 |
swaps out 'unused' pages as soon as possible to free memory and keep startup time
|
30 |
short for programs being loaded.
|
31 |
> |
32 |
>The behaviour of Linux is actually hotly debated by the developers as you can |
33 |
see on these two pages, which btw also include some hints on how to tune swapping:
|
34 |
> |
35 |
>http://kerneltrap.org/node/view/3000 |
36 |
>http://kerneltrap.org/node/view/3202 |
37 |
> |
38 |
>Have a nice weekend and kind regards |
39 |
>floki |
40 |
> |
41 |
> |
42 |
>-- |
43 |
>gentoo-performance@g.o mailing list |
44 |
> |
45 |
|
46 |
|
47 |
---- Introducing Spymac MailPro: http://www.spymac.com/mailpro/
|
48 |
|
49 |
--
|
50 |
gentoo-performance@g.o mailing list |