Technology news and Jobs 
VIRTUALISATION Enter ext4, the filesystem of the future
VIRTUALISATION Enter ext4, the filesystem of the future | Enter ext4, the filesystem of the future |
|
| by Sam Varghese | |
| Wednesday, 24 June 2009 | |
|
Page 3 of 3 iTWire: There are lots of good things said about zfs. What does it have that the options for Linux don't? Featured Whitepaper
5 Best Practices for Smartphone Support
Chris Samuel: Checksums on the filesystem data giving self-healing filesystems in the case of mirrors or RAIDZ or just error detection when you've just a single drive are, for me, the most interesting features. It also has the ability to snapshot the filesystem built in as well. You can run a Linux port of zfs under FUSE (originally done as a Google Summer of Code project) but don't expect good performance! I use it as the destination of rsyncs of my home directory which I then snapshot so I can go back in time should I accidentally delete something! There are emerging filesystems that offer this for Linux too, such as btrfs (checksums, snapshots) and NILFS (automatic checkpoints which can be converted into persistent snapshots before they get garbage collected) but they are both very new to the kernel and btrfs still does not have the disk format officially nailed down yet. iTWire: With Windows systems, people often need to defragment the drives. What causes this and why do we not hear of it with other operating systems? Chris Samuel: Disk fragmentation is, at its simplest, (a situation) where the file system cannot allocate contiguous areas of disk for file data. It gets spread out over the disk. Fragmentation causes issues because of the seek times on rotational drives; the file system has to wait for the disk heads to travel to the right place on the disk. Now of course SSD drives are meant to have much better seek times so it may be that this becomes less of an issue for them. Fragmentation happens for lots of reasons, but usually it's simply because you've deleted some files and so have freed up sections on the disk. Now if your disk is getting full then it might have to spread a larger file over multiple sections. FAT filesystems are very susceptible to fragmentation due to the very old design but all filesystems can suffer from fragmentation over time, though the major Linux filesytems are more intelligent about avoiding it. For instance the ext3 filesystem tries to mitigate fragmentation by preallocating extra blocks, just in case it needs to extend the file at a later date. However, it has no real defragmentation support built into the filesystem itself and relies on external tools being able to copy a file into a new single chunk before removing the old one. The XFS filesystem does, however, have a defragmentation tool called xfs_fsr (usually supplied in the xfsdump package for some reason) which will spot fragmented files and attempt to reallocate them into contiguous chunks of disk (i.e. making them into as few extents as possible). Of course, if you don't have enough free space to rewrite the entire file then it won't do as well! The new ext4 filesystem has been engineered with support for such defragmentation tools in the kernel, though the actual programs for sysadmins are still to be written. |
Tags
| < Next story in category | Previous story in the category > |
|---|
