Two Years With Linux BFS, The Brain Fuck Scheduler

Testing began with the Apache web-sever benchmark, which illustrated a difference in the original BFS benchmarking, and was still the case. When only one of the CPU cores was enabled, the Brain Fuck Scheduler came ahead of the Completely Fair Scheduler by about 5%, when two cores were enabled the CFS and BFS schedulers performed close to the same, and then when all four cores of the Core i5 2500K were enabled the Brain Fuck Scheduler was rather fucked. CFS was approximately 36% faster than BFS for this web-server benchmark on the quad-core Intel Sandy Bridge system. The Apache performance under BFS had only scaled by 9% where as the CFS was up by 46%.

While the results for Apache were rather interesting in comparing the BFS and CFS schedulers, using the NGINX web server was the opposite. This lightweight, high-performance web server was not impacted much by the switch in CPU schedulers, but it also had not scaled when going from the dual-core to quad-core configuration.

C-Ray is a very well multi-threaded ray-tracer, but using the Brain Fuck Scheduler was of no benefit on the Intel Sandy Bridge system. However, the system with the BFS scheduler was more responsive during the test process.