BFS Scheduler Benchmarks

The disk latency greatly increased under BFS. Carrying out eight threads of 16MB random writes resulted in the latency going from 23 microseconds under CFS to 33 seconds with BFS.

When looking at eight threads of 16MB reads (instead of random writes) the latency was still higher with BFS, but it was just 14.66 vs. 16.00 microseconds.

While BFS lost with the Threaded I/O Tester disk benchmark, it won with the PostMark test profile. The BFS scheduler delivered just fewer than 5% more transactions per second than the Completely Fair Scheduler.

Nine tests carried out on a single low-end system (we may carry out more tests on multiple systems if there is sufficient interest) is not enough to call one scheduler better than the other. However, based upon these numbers and the feedback from many others that have been testing out the Brain Fuck Scheduler, it does appear that BFS does have a sweet spot for most desktops (aside from the disk slowdowns). While we had not run into any stability issues, stalls, or other problems, others have encountered such issues with BFS being very young. The BFS scheduler does have a lot of potential as being the scheduler for Linux desktops, but at this time, Con Kolivas is not pursuing this for mainline integration. Should be change his mind or other kernel changes come because of his reappearance, we will be sure to keep you posted.

If you have tried out BFS, be sure to share your experiences in the Phoronix Forums.