EXT4 & Btrfs Regressions In Linux 2.6.36

With Compile Bench we have a look at where the performance fell between the Linux 2.6.34 and 2.6.35 kernels, but rather than being corrected in 2.6.36, the performance actually falls further. The EXT4 disk performance with Compile Bench barely changed between 2.6.35 and 2.6.36-rc3, but the Btrfs performance dropped by another 25%. Between the Linux 2.6.34 and 2.6.36-rc3 kernel releases the performance of Btrfs on the solid-state drive has dropped by nearly 40%.

The Flexible IO Tester performance only regressed slightly with the three tested kernels on the specified hardware.

With the Threaded I/O Tester when doing eight threads of 32MB random writes, the EXT4 file-system performance was maintained between Linux 2.6.34 and 2.6.36. Btrfs meanwhile dropped by 14% between Linux 2.6.34 and 2.6.35 and then between 2.6.35 and 2.6.36-rc3 it has dropped by an additional 11%.

These results are certainly a shock and not what we were expecting to see when testing the premiere Linux file-systems atop the latest kernel code that will be released as stable in just a month or two. The good news though is that these Linux file-system regressions do not appear across the board, but for example with our Intel Atom system with an HDD that is benchmarking the very latest kernel code on a daily basis at kernel-tracker.phoromatic.com don't suffer from these massive performance blows. Our investigation shall continue.

Discuss this article in our forums, IRC channel, or email the author. You can also follow our content via RSS and on social networks like Facebook, Identi.ca, and Twitter (@Phoronix and @MichaelLarabel). Subscribe to Phoronix Premium to view our content without advertisements, view entire articles on a single page, and experience other benefits.