A Gaggle of Schedulers in Kernel Development Battle

Sep 07, 2007

Really Fair - Really Simple, Really Fair - Really Unfair: three schedulers are the topic of current discussions on the kernel mailing list.

Developer Roman Zippel started the scheduler ball rolling by introducing his own version of the Completely Fair Scheduler (CFS), dubbed the "Really Fair Scheduler". CFS was developed by Ingo Molnar in collaboration with many other kernel developers, including Con Kolivas, who had quit kernel development just shortly before.

Referring to the Really Fair Scheduler presented by Zippel, Molnar said that it represented the next step of enhancements that could be introduced with kernel 2.6.24. However, the patch was not a fundamental change as it was based on the same mathematical principles and did not change the scheduler's behavior. Zippel was unfazed by this, emphasizing that the patch was an initial prototype, and pointing out the performance gains. What he had intended to do with the patch was to demonstrate changes to the algorithm. Molnar replied that it was unnecessary to write such a bulky patch; less code and less effort would have achieved the main changes.

A short time later, Ingo Molnar released his own version of Zippel's patch, which he dubbed "The Really Simple Really Fair Scheduler". In the mailing list Molnar commented that he had written the patch to better understand the mathematical principle behind Zippel's concept. The result was a simplified version. The name that Molnar gave it pokes fun at Zippel's patch. Zippel responded by stating that Molnar's patch was over-simplified. Molnar explained this by stating that he wanted to test what impact the scheduler would have on simple "nice-0-tasks" and how they would be sent to sleep and woken up again. According to Molnar, these processes account for 99 percent of all Linux processes. Molnar promises that he will revise the patch, however, he is not available for a couple of weeks.

The third scheduler to be introduced, "Completely Unfair Scheduling", is also based on CFS, but is in fact a daemon written in Python that helps to automate the process of optimizing the I/O and process schedulers. The Python script, which has now reached version five, and is actually called "Desktop Daemon" is designed to improve performance of games and audio applications. It supports both the "CFS process scheduler" and the "CFQ I/O scheduler", along with the "anticipatory I/O scheduler" and the "deadline I/O scheduler".

The script uses XML configuration files to define scheduler classes. Each of these classes includes its own set of scheduler optimizations. Individual applications or groups of applications can be assigned to groups. When one of the assigned applications is launched, the daemon automatically optimizes the individual scheduler settings for it.

Related content

  • Ingo Molnar Tests New BF Scheduler

    Kernel developer Ingo Molnar has done a benchmark test to compare his Completely Fair Scheduler (CFS) with the recently released BFS from Australian Con Kolivas.

  • Con Kolivas Introduces New BFS Scheduler

    After two years deep into Linux, the Australian Con Kolivas has emerged with a new scheduler that above all should provide significantly better performance on dual and quad processors.

  • Kernel News
  • Completely Fair Scheduler Analyzed

    Avinesh Kumar, an IBM developer has taken a close look at the Completely Fair Scheduler (CFS) introduced with kernel 2.6.23, comparing it with other schedulers. In the conclusion to the analysis, Kumar discusses changes in the future kernel 2.6.24 scheduler.

  • Realtime

    Linux provides tools and patches for speeding up the priority of multimedia applications. So if you're not getting the performance you expect, try shifting into overdrive.

comments powered by Disqus
Subscribe to our Linux Newsletters
Find Linux and Open Source Jobs
Subscribe to our ADMIN Newsletters

Support Our Work

Linux Magazine content is made possible with support from readers like you. Please consider contributing when you’ve found an article to be beneficial.

Learn More