need to make a fileystem for a cache used by our backup system. This cache will simply be 2 files of 2GB each.

My question is about the mkfs.ext3 parameters. I am assuming I need

1 - small number of inodes
2 - large block size
3 - large number of bytes per inode

Im thinking something along the lines of:

mkfs.ext3 -n -b 4096 -T largefile4 /dev/sda10 -v - block size of 4096 (max for ext3?) and 1 inode per 4MB (max for ext3?)

mke2fs 1.32 (09-Nov-2002)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
1472 inodes, 1502069 blocks
75103 blocks (5.00%) reserved for the super user
First data block=0
46 block groups
32768 blocks per group, 32768 fragments per group
32 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376, 294912, 819200, 884736



Anybody have anything to say on this im not a filesystem expert? Also what to do about journaling? Can it be tuned to have as little overhead on the perfomance as possible?

The end result just needs to be the best performance possible writing (when backing up) and reading (when restoring) for the 2GB files. Journaling is not important as far as recovering data goes as they can just be re-created if there is a problem they do not contain any static data.


Thanks

Hi

I'd use ext2 for that - it's a lot faster than ext3. Sorry, I'm no expert either - I think the default values for mkfs.ext2 would probably be fine.

Using the mount option "noatime" speeds reading time up quite a lot in my experience. With the default "atime", every time a file is read, the time is written to disk.

Some programs also put a lot of files in one directory. With some thousand files in a directory, it tends to slow down to a crawl. If that is the case, use ReiserFS instead, it doesn't slow down like ext2/3 does.