All should be well set to go ahead with migrating to ext4 filesystem, as stable version of ext4 is out with the latest linux kernel 2.6.28. Unlike ext2 to ext3, which was all about adding journaling feature to ext2, ext4 had lot of improvements.

Ext3 – 16 TB of maximum filesystem size, and 2 TB of maximum file size.
Number of subdirectories contained in a single directory in Ext3 is 32.000.

Ext4 adds 48-bit block addressing, so it will have 1 EB of maximum filesystem size and 16 TB of maximum file size. 1 EB = 1,048,576 TB (1 EB = 1024 PB, 1 PB = 1024 TB, 1 TB = 1024 GB).

Ext4 breaks that limit and allows a unlimited number of subdirectories.

Introduction of extends (Instead of indirect mapping of each and every blocks used by a file in ext3, ext4 produces ‘extents’ which says the data blocks of a file are in next ‘n’ blocks, mean a 200 MB file can be stored across a bunch of contiguous blocks ) in ext4 is set to bring a drastic performance upgrade and reduce fragmentation.

Unlike ext3, which can allocate only one block at a time, ext4 comes up with multi block allocator. (mballoc). This reduces disk allocation time greatly in ext4.

Delayed allocation functionality of ext4 will prevent excessive work done. In ext3, when a write function call is made for a 10MB file, it immediately starts allocating the 10MB even if the data is going t Delayed allocation functionality of ext4 will prevent excessive work done. In ext3, when a write function call is made for a 10MB file, it immediately starts allocating the 10MB even if the data is going to be in cache before it actually gets written to the disk. Where as ext4 only allocates disk space when actually the data gets written from cache to disk. This functionality works well with the above said extends and multiblock allocation features.o be in cache before it actually gets written to the disk. Where as ext4 only allocates disk space when actually the data gets written from cache to disk. This functionality works well with the above said extends and multiblock allocation features.

Next amazing thing about ext4 is the reduced fsck time needed ~ 2 to 20 times… whoo… This is made possible by writing a list of unused inodes at the end of each inode group table, so fsck can skip those inodes from checking. The free inode list is created by fsck when its ran first at the first time on the filesystem, from the very next fsck you will have the advantage of reduced checking time.

Journal checksumming is another important feature introduced in ext4, which writes and stores checksums of filesystem journal. This speeds up ext4 20% compared to its predecessor ext3 by omitting two phase commit system.

Online defragmentation is set to be live in future releases of linux kernel which supports a tool named e4defrag, which do defragmentation of ext4 filesystems.

inode related features,

* Larger inodes for accommodating nano time stamps, inode reservation.
* nano timestamps for files
* Inode reservation, when a directory is created some inodes are created in advance for faster performance.
* faster extended attributes

Persistent preallocation

Barriers on by default

Migrating from ext3 to ext4

You need to use the tune2fs and fsck tools in the filesystem, and that filesystem needs to be unmounted. Run:

tune2fs -O extents,uninit_bg,dir_index /dev/yourfilesystem