As in new Linux 2.6.28 kernel ext4 file-system [http://en.wikipedia.org/wiki/Ext4] marked as stable it is time to decide if there is any sense to reformat our hard disks.
Compared performance of the EXT3, EXT4, XFS, and ReiserFS file-systems.
Dual Intel Xeon E5320 Quad-Core “Clovertown” processors
Tyan Tempest i5400XT motherboard
2GB of Kingston DDR2 FB-DIMM RAM
160GB Western Digital WD1600YS-01SHB1 hard drive
ASUS GeForce 9600GT 512MB graphics card.
Standard installation of Fedora 10 x86_64, without Logical Volume Manager
constant SWAP size: 4GB
The Serial ATA hard drive was using AHCI mode.
operations per second on following graphs:
Here the number of sustained operations per second had dropped about 15% when changing from EXT3 to EXT4. The fastest file-system was XFS followed by the ReiserFS with a slight lead over EXT3.
The benefits of EXT4 were shown in a 4GB sequential create. EXT4 managed to push a 34% lead over EXT3 and sizable leads over XFS and ReiserFS.
EXT4 had continued to shine through with the 4GB sequential read. XFS managed to outperform EXT4 at reading, but it was by a very slim margin. EXT3 and ReiserFS were nearly 40% slower.
And then some real examples:
To see what difference the file-system makes when encoding media files, we had used LAME for WAV to MP3 conversion and FFmpeg for AVI to NTSC VCD. When encoding our sample WAV file using LAME, all four file-systems took 52 seconds with first and last place being just a fifth of a second difference.
If you are interested in encrypting files with GnuPG, the fastest 2GB file encryption occurred with EXT3 and XFS in a close second. This was followed by ReiserFS and then EXT4.
While the LAME results were incredibly close, there was a greater spread when looking at our FFmpeg numbers. Here the fastest file-system was EXT3, which came out to be about 6% faster than EXT4. ReiserFS had finished with a close second followed by XFS in third.
EXT4 introduces Extents to replace the block mapping scheme used on earlier generations of the extended file-system, persistent pre-allocation, delayed allocation, precision timestamps, and journal check summing.
EXT4 is more scalable, more efficient through the use of Extents, supports larger disk capacities, can handle twice the number of sub-directories, is capable of handling online defragmentation[http://en.wikipedia.org/wiki/Defragmentation], and there is improved reliability via journal checksums. What perhaps is more important is that with the addition of these new features, the performance hasn’t regressed. Also, when testing the EXT4 file-system, we hadn’t run into any problems with stability, file corruption, or any other issues.
based on phoronix results: