1 |
Mark Knecht schrieb: |
2 |
> Hi, |
3 |
> I'm wondering if anyone has opinions (on this list? Right...) as to |
4 |
> the best file system type for running vmware images of Windows XP. As |
5 |
> best I can tell an 8GB C: drive shows up as 4x 2GB files and 1.5GB |
6 |
> DRAM is modeled in a file of its own: |
7 |
> |
8 |
[...]I suspect XFS, Reiser or |
9 |
> something other than ext3 that I'm currently using would possibly make |
10 |
> a difference? |
11 |
> |
12 |
> But would it be a big difference? How would I test it? I.e. - not |
13 |
> that XFS is or isn't a good file system but that it actually helps |
14 |
> with vmware operation? |
15 |
> |
16 |
> My current system is AMD64, 4GB DRAM, i5 Core which is 4 processor |
17 |
> threads. Target machine is AMD64, i7 Core or possibly dual XEON, 16GB, |
18 |
> 4-6 monitors and multiple 1TB+ drives in some sort of RAID mainly for |
19 |
> speed, and then additionally an external RAID for backup. |
20 |
> |
21 |
|
22 |
As an educated guess, I would say that ext4's usage of extents should |
23 |
improve the performance on large files by a measurable margin. |
24 |
|
25 |
As one of the last filesystem drivers which still use the Big Kernel |
26 |
Lock, I would also guess that Reiserfs might not scale well to heavily |
27 |
multithreaded usage or multicore CPUs. However, I cannot say if this has |
28 |
any kind of noticable effect on 4-8 core machines. |
29 |
|
30 |
As Roberto Waltman already suggested, benchmarking is the way to go. |
31 |
|
32 |
|
33 |
By the way, as you mention RAID. Note that on pure read operations, a |
34 |
RAID1 can be as good or even better than a RAID0 because read operations |
35 |
can be distributed freely among the disks since all of them contain all |
36 |
data. |
37 |
|
38 |
Hope this helps |
39 |
Florian Philipp |