[arch-general] High fragmentation rate with ext4?
Hello everyone! As linux 2.6.30 is out, I decided to migrate my data partition to ext4. Hoping for better performance and less fragmentation. But, after format it with "mkfs.ext4 - largefile /dev/..." and copy all the file back to the partition, I got whooping number of 30% non-contiguous files! Then I reformat it with reserved block set to 10% (-m 10), still got the same fragment rate :-( Okay, so I try it again, this time with ext3, then I do a "mount -a" to remount all the partitions, copy some files (size between 2 MBs~40 MBs) back in, guess what? This is the result of fsck: 81 inodes used (0.07%) 69 non-contiguous files (85.2%) Then I remember that "mount -a" mount the partition as ext4 (since I haven't changed my fstab yet), okay, so I format it again, and mount it with "mount -t ext3 ....", copy some files in and this is the result: 81 inodes used (0.07%) 9 non-contiguous files (11.1%) Still not as good as it was before, but much better than the 85.2 rate above. So is this a bug? Does anyone have this problem? My e2fsprogs version is 1.41.5-2
2009/6/11 Cá Vàng <clarious@gmail.com>:
Hello everyone!
As linux 2.6.30 is out, I decided to migrate my data partition to ext4. Hoping for better performance and less fragmentation.
But, after format it with "mkfs.ext4 - largefile /dev/..." and copy all the file back to the partition, I got whooping number of 30% non-contiguous files! Then I reformat it with reserved block set to 10% (-m 10), still got the same fragment rate :-(
Some updates, I tested copy one big file (about 3.6 GBs) to the new partition, and this is what I get: (ext3 mount as ext3)image.iso: 31 extents found, perfection would be 29 extents (ext3 mount as ext4)image.iso: 960 extents found, perfection would be 29 extents -- My job: Writing buggy softwares to make the world a worse place for humanity ( ゚∀゚)アハハ八八ノヽノヽノヽノ \ / \/ \
participants (1)
-
Cá Vàng