1 |
On Mittwoch, 13. Februar 2008, Richard Freeman wrote: |
2 |
> Volker Armin Hemmann wrote: |
3 |
> > On Mittwoch, 13. Februar 2008, Richard Freeman wrote: |
4 |
> >> Do you have benchmarks to support this? |
5 |
> > |
6 |
> > which numbers? that swap is horrible slow compared to ram? |
7 |
> |
8 |
> No - that compiling in tmpfs is horrible compared to compiling to disk. |
9 |
> Neither of us is proposing getting rid of disk access entirely - we're |
10 |
> just debating two different ways of doing it. |
11 |
|
12 |
tmpfs is fine - if you have enough ram. 1gb is not enough. |
13 |
|
14 |
|
15 |
|
16 |
> >> As opposed to wasting all your RAM on cache&buffers to hold all those |
17 |
> >> files being accessed? During compilation that RAM is going to be |
18 |
> >> heavily used - no getting around that. Once you're done any files left |
19 |
> >> sitting on a tmpfs will just get paged out until accessed. They |
20 |
> >> shouldn't really use any RAM at all. Even if those files were on disk |
21 |
> >> they would consume RAM in the form of caching until they're considered |
22 |
> >> unneeded. |
23 |
> > |
24 |
> > cached&buffered information can be discarded anytime if the ram is needed |
25 |
> > elsewhere. tmpfs has to be shoved into swap. |
26 |
> |
27 |
> Uh, data written to a file has to be written to disk before the |
28 |
> cache/buffer is flushed. The difference is that the data is written to |
29 |
> disk within about 10 seconds of being written to the buffer, and then |
30 |
> retained in cache maybe for a few hours longer (optimistically). With |
31 |
> tmpfs the write doesn't happen until the space is needed. Either way |
32 |
> every byte gets written - with tmpfs it is deferred as late as possible |
33 |
> but with disk-based filesystems the write is done as early as possible. |
34 |
> |
35 |
|
36 |
emm, no. |
37 |
|
38 |
|
39 |
> > Extreme example? |
40 |
> > |
41 |
> > kdepim with the enablefinal flag. On amd64 every single gcc instance |
42 |
> > needs ~900mb at two points of compilation. With 1gb ram and no tmpfs it |
43 |
> > sucks. |
44 |
> > |
45 |
> > With 1gb ram and 512mb of that reserved for tmpfs, you'll get a |
46 |
> > swapstorm. |
47 |
> |
48 |
> Oh, it will certainly swap a great deal. However, I don't see why it |
49 |
> would be any slower - either way you're doing a ton of disk access. Do |
50 |
> you have actual benchmarks? |
51 |
|
52 |
Not real benchmarks. But kdepim with enablefinal, 1gb of ram and -j2 several |
53 |
hours. With j1 2h. |
54 |
. |
55 |
kdepim with 2gb of ram and j2 30m |
56 |
Just because first case swap storn, last case no swap at all. |
57 |
|
58 |
> |
59 |
> Again, tmpfs doesn't "reserve" memory - it uses memory - just like |
60 |
> cache/buffers. |
61 |
|
62 |
but while cache/buffers can be discarded when the ram is needed, tmpfs has to |
63 |
be shoved into butt-slow swap. |
64 |
|
65 |
> |
66 |
> >> I certainly agree that |
67 |
> >> swap is slow compared to RAM, but it isn't slow compared to a disk-based |
68 |
> >> filesystem. |
69 |
|
70 |
yes, yes it is. It is faster to start an app from disk, then to fetch it out |
71 |
of swap. My very personal experience since many many years. |
72 |
|
73 |
|
74 |
> > |
75 |
> > really? Every really swapped? IMHO it feels like every single byte is |
76 |
> > fetched by a mule. |
77 |
> |
78 |
> Right now I've got 589MB of free RAM (-/+ buffers/cache) with 2GB of RAM |
79 |
> total. I've got 705MB of swap used. Works just fine IMHO. Sure, if I |
80 |
> make it really busy it can get slow, although with nice/ionice there |
81 |
> isn't much visible |
82 |
|
83 |
your system would feel and act a lot faster if you don't have anything in |
84 |
swap. 'Fine' is good - as long as you don't know the alternative. |
85 |
|
86 |
> > Of course adding more RAM will never hurt, but that incurs |
87 |
> > |
88 |
> >> significant cost and you can at least maximize your current hardware |
89 |
> >> before investing in more of it. |
90 |
> > |
91 |
> > significant costs like 30euros for 2gb? |
92 |
> |
93 |
> Sure. Compared to about $1 for 2GB of hard drive space 30 euros is |
94 |
> significant. |
95 |
|
96 |
don't forget that ram is also roughly 100 times faster than harddisk. |
97 |
|
98 |
> And even if he had 2GB of RAM I submit that it would |
99 |
> probably work better if he compiled in tmpfs than on disk. |
100 |
|
101 |
with 2gb he could do it. But with 1gb he just hurts himself. |
102 |
|
103 |
|
104 |
> There is no |
105 |
> question that a $550 computer will outperform a $500 computer, and a |
106 |
> $600 computer will outperform a $550 computer, and so on. You can |
107 |
> always spend 30 more euros and improve your system. What I'm interested |
108 |
> in is maximizing the performance of the system I have now. I can always |
109 |
> spend $50 and make it even faster. However, there are lots of things I |
110 |
> can spend $50 on - I'd rather spend it on something else all things |
111 |
> being equal. |
112 |
|
113 |
you could stop shoving everything in swap - no costs involved and system is a |
114 |
lot faster. |
115 |
|
116 |
|
117 |
-- |
118 |
gentoo-amd64@l.g.o mailing list |