I found this thread from a few months ago that seems to have been started by someone with the same problem I have, but it was not resolved: https://www.redhat.com/archives/ext3-users/2005-March/msg00000.html I have been trying to speed up my boot times and I notice that a lot of time is spent waiting on IO. Even with readahead-list precacheing files, it is reading at less than 1/4 of my disk's maximum sequential throughput. I feel this is because the files are scattered all around the disk, which causes a lot of seeking. I am looking for a way to defragment the disk such that the files needed for boot are all packed sequentially at the beginning of the disk, so that readahead-list can read them in much faster. During that previous thread, it was stated that it was impossible to have defrag work safely on an ext3 partition. I do not see why this is so. If ext2 could be given a journal to make it crash resistant, then why can't defrag be given the same thing? Another comment was made saying that such a defragger could not possibly be done only in userspace. Unless you're talking about online defragmentation, then manipulating the filesystem offline to repack it is entirely possible from userspace. Is there still no way to do what I am trying? P.S. Please CC me on replies, as I am not subscribed.
-----BEGIN PGP SIGNED MESSAGE----- Hash: RIPEMD160 Phillip Susi schrieb:> I am looking for a way to defragment the disk such that the files needed > for boot are all packed sequentially at the beginning of the disk, so > that readahead-list can read them in much faster.i've never used ext2fs defrag-programs so i don't know how "safe" they are. but if they're working on ext2 - why not tune2fs' an existing ext3 into a ext2, run the defrag program and tune2fs' into ext3 again: % umount /dev/sdX % tune2fs -O ^has_journal /dev/sdX % <run magic defrag program on sdX> % tune2fs -O has_journal /dev/sdX % mount /dev/sdX obviously this has to be done while the device is offline and i'd recommend having backups of course. i've never tried it though... - -- BOFH excuse #86: Runt packets -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDpFxeC/PVm5+NVoYRA4yYAKC8q6yKJHWaO54bq/97j6P8n4+GtgCfXE0O lmTPqGcJINRW0zNjaxYdpOI=iFbD -----END PGP SIGNATURE-----
Possibly Parallel Threads
- Repacking database from v1 to v2 format: how long does it take?
- Repacking database from v1 to v2 format: how long does it take?
- Repacking database from v1 to v2 format: how long does it take?
- Repacking database from v1 to v2 format: how long does it take?
- FYI: Why is NFS slower on EL5 than EL4?