which file system to use

[Alexander Skwar]

David Abrahams <dave at boost-consulting.com>:

> Alexander Skwar <listen at alexander.skwar.name> writes:
> 
>> John L Fjellstad <john-ubuntu at fjellstad.org>:
>>
>>> Alexander Skwar <listen at alexander.skwar.name> writes:
>>> 
>>>> The commit is not (only) because of atime. ext3 will commit to
>>>> disk, even if *nothing* has been accessed.
>>>>
>>>> That said, laptop-mode is a good suggestion.
>>> 
>>> Looking at the mount options for ext3, there is a commit option
>>> (although the default is set to 5 sec, so I'm not sure it's that which is
>>> causing the disk access).  If it is the cause of the disk access, I
>>> would think changing this option to something else (60 sec or whatever),
>>> might help.
>>
>> Well - turning on/off the hd every 60 seconds isn't *THAT* much
>> better, is it? ;) I'd rather set it to 1 hour, or so. BUT: If
>> the system crashes, you'll lose the not yet written data. Ie.
>> up to 1 hour would be lost.
>>
>> Because of this and because of the slowness, I'd not suggest ext3
>> but rather JFS.
> 
> [I have a couple of machines (a laptop and a server) with fresh
> installs on ext3, and at least one of them probably needs a full
> reconfiguration, so I'm following this thread with great interest.]
> 
> Two questions:
> 1. Does JFS always commit immediately?

I don't know for sure, but I don't think, it'll commit *immediately*.
This would be bad performancewise, I'd think. I'd suspect, that it,
let's say, "groups" together requests, so that the disk access can
be minimized.

> 2. What is the purported purpose of the ext3 commit delay?

No clue.

Alexander Skwar
-- 
BTW: I have a better name for the software .... Microsoft Internet
Exploder.
        -- George Bonser <grep at cris.com>