>>>>> "Joseph" == Joseph A Knapka <jknapka@earthlink.net> writes: Joseph> Chris Mesterharm wrote: >> >> Hello, >> >> I was hoping someone could suggest an area within the Linux kernel >> that could benefit from research in machine learning. Joseph> People have suggested VM, but frankly, I think the issues with Joseph> the Linux VM subsystem are fairly well-understood, and the main [...] Joseph> However, one place where ML could, IMO, have a significant Joseph> impact, is on the page-replacement policy. Any method that Strange enough, but people talking about "VM" actually have in mind page replacement indeed. Joseph> lets us make more accurate choices about what pages to read Joseph> in from disk and what pages to throw away, will be a win, Joseph> because it will postpone the thrash point. For example, if Joseph> you have an algorithm that, over many runs of a program, makes Joseph> weighted associations between the program inode and text Joseph> and data pages that it uses, you may be able to make very Joseph> accurate page-replacement choices *without* having to let Joseph> pages filter all the way through the page cache. In other Joseph> words, you may be able to implement a policy much closer Joseph> to "Not Going To Be Used" than the current "Not Recently Used". Joseph> That could significantly improve VM performance, and may Joseph> have a positive influence on filesystem performance as Joseph> well. A suggestion/gut feeling/vague idea I had is to make linux recognize memory access patterns of processes as a base for a prediction of the future paging behavior. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ IRC Channel: irc.openprojects.net / #kernelnewbies Web Page: http://www.kernelnewbies.org/
