For those thinking of playing with predictive caching (likely an area of considerable student endeveour/interest these days at both filesystem and "web" level):
--- Matthew Dillon: > So there is no 'perfect' caching algorithm. There > are simply too many variables even in a well defined > environment for even the best system heuristics to > cover optimally. --- David Schultz: > If that proves to be infeasible, I'm sure there are > ways to approximate the same thing. The hard parts, > I think, would be teaching the VM system to use the > new information, and gathering statistics from which > you form your hints. --- Right. It's easy if you know the complete future of the total system state, which of course you never will. Someone interested in this might try to apply the latest in machine learing techniques, classifiers, etc., to the online problem. Variants of this are receiving lots of attention in areas such as gene sequence prediction. I dunno, but it seems like a lot of the math ends up pretty similar to economics, and we all know how well those models work. Kind of funny, running an economic simulation in your kernel... but actually getting possible at some level, at least for research systems with modern machines. There was a time when you would be fired for putting floating-point in an OS. ---- http://csl.cse.ucsc.edu/acme.shtml : "Many cache replacement policies have been invented and some perform better than others under certain workload and network-topological conditions. It is impossible and sub-optimal to manually choose cache replacement policies for workloads and topologies that are under continuous change. We use machine learning algorithms to automatically select the best current policy or mixtures of policies from a policy (a.k.a expert) pool to provide an "adaptive caching" service." - bruce To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
