Can DragonFlyBSD's HAMMER Compete With Btrfs, ZFS?

The disk read performance with BlogBench did the worst by far with the HAMMER file-system. When maintaining the same DragonFlyBSD 2.8.2 stack but switching from HAMMER to the UFS file-system, the performance was actually improved by nearly five times. Interestingly though, UFS on PC-BSD 8.1 was much faster than UFS on DragonFlyBSD either due to differences in the UFS implementation or with the disk/SATA sub-system between these distinct BSD operating systems. Like the HAMMER results, the performance of PC-BSD with a root ZFS file-system was actually slower than using UFS. Meanwhile, all of the tested Linux file-systems were much faster than ZFS and HAMMER on BSD. The performance of EXT4 and Btrfs were neck-and-neck.

While HAMMER did not have any advantages over UFS when it came to the read performance in BlogBench, the HAMMER write performance in this test profile was much faster than the classic BSD file-system. The UFS performance on DragonFlyBSD continued to be slower than UFS on PC-BSD, but HAMMER was able to outperform UFS on PC-BSD and even ZFS on PC-BSD, which itself was slower than UFS. BSD and HAMMER though was still no match for the EXT3/EXT4/Btrfs that all performed much stronger. Interestingly though, with this test the Btrfs performance fell greatly behind EXT4.

The Gzip compression test is not too interesting, but Linux and its file-systems continue to command the lead.

With the PostMark test, HAMMER on DragonFlyBSD was five times faster than UFS on the same operating system. The performance of HAMMER under DragonFlyBSD though fell short of ZFS on PC-BSD. The ZFS performance in turn was not too far behind EXT3 on Linux, but EXT4 and Btrfs were both multiple times faster.