1 |
Peter Humphrey posted <200605020855.21021.prh@××××××××××.uk>, excerpted |
2 |
below, on Tue, 02 May 2006 08:55:20 +0100: |
3 |
|
4 |
> On Monday 01 May 2006 13:09, Duncan wrote: |
5 |
> |
6 |
>> All that said, as I've mentioned a couple times now, on a single disk |
7 |
>> system, there's a /huge/ cost in terms of access speed, to access swap. |
8 |
> |
9 |
> So, following a suggestion of yours some time ago, as I have 4 GB RAM in |
10 |
> this box I compiled the kernel without swap code and ran without a swap |
11 |
> partition. It seemed to work well, though I didn't run any proper tests. |
12 |
> I've reverted to using swap more recently because I'm running the new |
13 |
> vmware-server package, which seems to eat memory wholesale and I don't |
14 |
> want to risk running out of space. I don't think swap space is actually |
15 |
> being used though. So far. |
16 |
|
17 |
Now that I've 8 gig of physical memory, I've seriously debated turning off |
18 |
swap again. The 4x4 gig striped swap is fast enough that using it isn't a |
19 |
problem. However, the suggestion you mention, that the swap code in the |
20 |
kernel is a whole section of complex code that can be disabled and removed |
21 |
from the kernel entirely if swap is disabled at compile time, is still |
22 |
100% valid. I haven't decided whether the small chance of actually using |
23 |
swap justifies keeping swap in the kernel or not. I suppose it doesn't, |
24 |
and I'd certainly remove it if I didn't have the 4-way-striped swap, but |
25 |
with the striped swap, I haven't decided. |
26 |
|
27 |
Actually, the reason I re-enabled it again was to try swsusp, which writes |
28 |
the memory image to the swap partition, so there has to be such a thing |
29 |
and AFAIK swap has to be enabled, too. However, I've had decidedly mixed |
30 |
results. Back with 2.6.15 (the last I tried), anyway, it seemed I /could/ |
31 |
suspend if I'd not activated either the network or X. If I'd activated |
32 |
one or either and then quit X and stopped the network, it would sometimes |
33 |
work, sometimes not. |
34 |
|
35 |
I should really try again with .17-rcX and the 8 gig memory, which I |
36 |
didn't have back then either. If I can get it working reliably, great, |
37 |
that's a good reason to keep swap enabled. If not, I should kill the |
38 |
entire supporting infrastructure in the kernel, swsusp, CPU-hotplug |
39 |
(necessary for swsusp on a dual CPU system), and swap as well. |
40 |
|
41 |
The thing is finding the time and motivation to do it. It takes quite |
42 |
some patience, since testing it involves shutting the hardware down, and |
43 |
seeing if it can successfully resume when I turn it back on. If it |
44 |
doesn't, that's a full boot sequence. Even if not, it's a full BIOS |
45 |
power-on sequence, and this BIOS has one VERY annoying |
46 |
bug/feature/trait -- it insists on checking the floppy (and cd and dvd, |
47 |
but floppy is the worst) on boot, even if I have the BIOS set to NOT check |
48 |
it (and not boot from it). That adds about 30 seconds to the boot |
49 |
sequence right there, while the thing clicks away on the floppy. (This |
50 |
particular BIOS is on the Tyan site as a beta, not full release, but it |
51 |
fixed a couple other things, so I wanted/needed it, even with this |
52 |
irritating bug.) |
53 |
|
54 |
-- |
55 |
Duncan - List replies preferred. No HTML msgs. |
56 |
"Every nonfree program has a lord, a master -- |
57 |
and if you use the program, he is your master." Richard Stallman in |
58 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
59 |
|
60 |
|
61 |
-- |
62 |
gentoo-amd64@g.o mailing list |