[rescue] $35 50gig drives

Sean Batt sean at coombs.anu.edu.au
Tue May 10 20:52:30 CDT 2005


On 11/05/2005, at 10:26 AM, Joshua Boyd wrote:
> On Tue, May 10, 2005 at 10:40:01PM +0000, Peter Corlett wrote:
>> Joshua Boyd <jdboyd at jdboyd.net> wrote:
>> [...]
>>> I did not know that. I guess there isn't much point to using JFFS2
>>> on a CF device then.
>> Actually, a journalling filesystem is liable to wear the card out much
>> faster than, say, ext2. You will probably want to mount it noatime if
>> you're mounting it read-write.
>
> You should look up the paper on JFFS2.
> http://sources.redhat.com/jffs2/jffs2.pdf

Normally I think its a great idea for people to read up on all the  
filesystems
available, proposed, historic or based on perfect knowledge. Normally...

But ext2 doesn't have to write to the "disc" at all, its happy to keep
lots of modified data in memory. Ext2 wants to be fast for the general
case and not worry about on-disc consistency in the face of a power
failure (etc) which is what you get with journaled filesystems.

123456789012345678901234567890123456789012345678901234567890123456789012 
34567890

I've seen a graph of number-of-disc-writes for a variety of file systems
compiling the linux kernel and ext2 writes much less than the journals,  
the
softly updated, the synchronous, etc. (Can't seem to locate the graph at
the moment, of course; today Google isn't my friend).

Sean
--
Sean.Batt at anu.edu.au



More information about the rescue mailing list