1 |
On Wednesday 06 May 2009 19:39:02 Duncan wrote: |
2 |
snip |
3 |
snip |
4 |
snip |
5 |
> Third, tmpfs is useful in that it isn't restricted to physical memory, |
6 |
> and can use swap as well, if there's memory pressure and something in |
7 |
> tmpfs to swap out. Thus the "worst case" as mentioned earlier, that |
8 |
> there's not enough memory and the system has to write the files to disk |
9 |
> (swap, now) after all. But another implication, then, is that swap size |
10 |
> matters too. It can't use swap you don't have configured and mounted! |
11 |
> The old rule of thumb was to have swap of twice the size of regular |
12 |
> memory. While that no longer really applies as it used to, because if |
13 |
> someone has memory of say four gigs, it's unlikely they're going to be |
14 |
> prepared to wait for 8 gigs of swap to fill up even if they have it |
15 |
> configured (at least if they're running a single disk, RAID-0/striped |
16 |
> swap is faster and thus not as bad), for those with a gig of RAM or more, |
17 |
> I'd say 2 gigs swap minimum is reasonable. Those using tmpfs for |
18 |
> anything major, however, as we're talking about here, will probably want |
19 |
> more, say 4-6 gigs swap, just in case. |
20 |
|
21 |
Just as a note swap is striped by kernel if you set the individual swap |
22 |
partitions to same priority level. So no need to build swap partiotion as an |
23 |
RAID array. |
24 |
|
25 |
I use RAID 1 (mirrored) swap partition of 4GB in size just to protect swap |
26 |
outs from disk failures. |
27 |
I simply want this system to be a more resilient to disk failures. :) |
28 |
|
29 |
snip |
30 |
snip |
31 |
snip |
32 |
sharpens the scissors :) |
33 |
snip |
34 |
snip |
35 |
|
36 |
> OTOH, while multiple tmpfs mounts increases overall resource exposure, |
37 |
> it /does/ allow one to better restrict access to individual tmpfs |
38 |
> mounts. Perhaps that's Sami's strategy. If he limits writing (and |
39 |
> perhaps reading too) on his PM temp to the portage user (and root, of |
40 |
> course), then by separating them, he's limiting exposure on his (assumed |
41 |
> world writable) 2-gig system tmpfs to its 2-gigs, while I'm exposing that |
42 |
> whole 6-gig tmpfs to writes by any user. As I said, perhaps he'll post |
43 |
> his reasoning and we'll see. |
44 |
|
45 |
I simply mount my PM temp to /var/tmp/paludis and chown it as |
46 |
paludisbuild:paludisbuild to restrict this 5GB space to be only for paludis |
47 |
temp usage. |
48 |
|
49 |
The system /tmp I restricted to 2GB as I some times need more tmp space than |
50 |
my earlier restriction, which was about 800MB. |
51 |
|
52 |
The reason I restrict my /tmp quite small is cache. This way I'm not tempted |
53 |
to use the /tmp too much. The reason is that I simply want to allow the |
54 |
system to keep almost every program and most of the files I use in the cache |
55 |
memory. This will make the system more responsive. Besides I have sometimes |
56 |
forgot that I had something in my /tmp and after a reboot... I of course did |
57 |
some things again because I like that so much. :) |
58 |
|
59 |
The only thing that anoyingly invalidates my io buffers is listening music. |
60 |
Maybe I should cut the size of my Music library heavily. :) |
61 |
|
62 |
The real reason why I use tmpfs for the PM temp dir is not that it makes |
63 |
compiles so much faster, but the fact that it will also ease the unnecessary |
64 |
disk writes that building packages generate. This will result in a longer life |
65 |
time for my disks. |
66 |
|
67 |
snip |
68 |
snip |
69 |
OK Does anyone want to buy a "slightly" used scissors? :) |