1 |
On 8/29/06, Ow Mun Heng <Ow.Mun.Heng@×××.com> wrote: |
2 |
> On Tue, 2006-08-29 at 18:35 -0700, Richard Fish wrote: |
3 |
> > On 8/29/06, Ow Mun Heng <Ow.Mun.Heng@×××.com> wrote: |
4 |
> > > $ mount | grep xfs |
5 |
> > > /dev/hda6 on /home type xfs (rw) |
6 |
> > |
7 |
> > Hmm, I missed this before. "nobarrier" should be showing up here. Try: |
8 |
> > |
9 |
> > mount /home -o remount,nobarrier |
10 |
> > |
11 |
> |
12 |
> I did mention that I tried that as well. (but I just re-tried it anyway) |
13 |
> and it didn't have any changes. |
14 |
|
15 |
I don't think it's a good idea. Do you know what the write barriers |
16 |
provide? They give you a much higher chance of no damage if you are |
17 |
able ot run with barriers. The only real danger to an XFS partition |
18 |
is out-of-order commits (and of course massive hardware failure). |
19 |
Write barriers prevent out-of-order journal vs. FS commits, and that |
20 |
is a Good Thing(tm). |
21 |
|
22 |
|
23 |
They are only in as of 2.6.17 for XFS. |
24 |
|
25 |
> |
26 |
> Out of curiousity, I just tried to copy a file using an xterm (instead |
27 |
> of using nautilus) from DIsk 2 to disk1 |
28 |
> |
29 |
> disk2/partition2 - VFAT |
30 |
> disk1/partition6 - XFS partition (/home) |
31 |
> |
32 |
> |
33 |
> $ ls -lah WinXP-000001-cl1-000001-cl1-000001-s002.vmdk |
34 |
> -rwxrwxrwx 1 root root 620M Aug 29 19:10 |
35 |
> WinXP-000001-cl1-000001-cl1-000001-s002.vmdk |
36 |
> |
37 |
> $ time cp WinXP-000001-cl1-000001-cl1-000001-s001.vmdk ~/Desktop/ |
38 |
> real 0m37.353s |
39 |
> user 0m0.157s |
40 |
> sys 0m5.445s |
41 |
> |
42 |
> |
43 |
> Transfer rate ~16.8MB/s |
44 |
|
45 |
I read an article recently (<2 years ago , heh) about how Nautilus, |
46 |
mc, Cp and other OSS copy utilities suffer from a lack of real |
47 |
intuitive optimization work. They highlighted the GNU cp command. |
48 |
Overall copying with Linux tends to be lack luster. It's services we |
49 |
do well. i.e. Http, SQL, etc. So, that begs the question: Are there |
50 |
any good, well optimized file copy utilities for Linux? |
51 |
|
52 |
> |
53 |
> Using Nautilus (I don't know of a good way to measure throughput using |
54 |
> this, so it's basically what I see in the progress bar |
55 |
> ~5min |
56 |
> gkrellm2 notes transfer rate ~2.0MB/s |
57 |
> |
58 |
> |
59 |
> However, doing the same thing to my /tmp directory (ext3 partition) the |
60 |
> same file copies in ~30secs and w/ ~17MB/s transfer rate. |
61 |
> |
62 |
> What gives?? |
63 |
> |
64 |
> |
65 |
> BTW, what's the difference between mc and mc-mp?? |
66 |
> |
67 |
> * app-misc/mc |
68 |
> Latest version available: 4.6.1 |
69 |
> Latest version installed: [ Not Installed ] |
70 |
> Size of downloaded files: 11,606 kB |
71 |
> Homepage: http://www.ibiblio.org/mc/ |
72 |
> Description: GNU Midnight Commander cli-based file manager |
73 |
> License: GPL-2 |
74 |
> |
75 |
> * app-misc/mc-mp [ Masked ] |
76 |
> Latest version available: 4.1.40_pre9 |
77 |
> Latest version installed: [ Not Installed ] |
78 |
> Size of downloaded files: 2,904 kB |
79 |
> Homepage: http://mc.linuxinside.com/cgi-bin/dir.cgi |
80 |
> Description: GNU Midnight Commander cli-based file manager. 4.1.x |
81 |
> branch |
82 |
> License: GPL-2 |
83 |
|
84 |
>From the URL: |
85 |
"The goal of this project is creating a stable, well-working, usefull |
86 |
console-only version of well-known Midnight Commander, without bugs |
87 |
and garbage, like tk, xv and gnome. I'm bored waiting for bugfixes, |
88 |
and A'rpi's ESP team stops their work in this direction too, so I did |
89 |
it. I'm fixing all (found) bugs, reported by my friends, and made some |
90 |
really pleasent new features, like real-time clock, or filegroups |
91 |
colorizing." |
92 |
|
93 |
Basically, this guy is sane, (thank God), and doing what MC really |
94 |
needs: SIMPLICITY. I have never bitched at the mc guys, but when I |
95 |
get to my desk, I am converting to mc-mp. Cause MC sucks recently. |
96 |
-- |
97 |
gentoo-user@g.o mailing list |