tag:blogger.com,1999:blog-5799246.post6358462845663808948..comments2023-07-17T17:51:20.831+00:00Comments on a very occasional diary @ Nikita Danilov: File system replacement algorithms (cont.)nikitahttp://www.blogger.com/profile/09403336533089968821noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-5799246.post-14809166710887284172007-07-03T13:31:00.000+00:002007-07-03T13:31:00.000+00:00A friend of mine co-owns a heavily loaded shell/we...A friend of mine co-owns a heavily loaded shell/web/mail hosting machine. It might be quite interesting to gather some data when it's under heavy load. That happens quite often, and the poor machine is swapping like mad. But tracing a production machine is a difficult bussiness...<BR/><BR/>I just started to look at ZFS which uses ARC, and I'm surprised by the drop in ARC results at 2GB.<BR/><BR/>BTW, will there ever be more reiser4 articles?Unknownhttps://www.blogger.com/profile/03087832713852077541noreply@blogger.comtag:blogger.com,1999:blog-5799246.post-14398134743995621532007-06-01T17:49:00.000+00:002007-06-01T17:49:00.000+00:00I wonder how well a Most Recently Used (MRU) algor...I wonder how well a Most Recently Used (MRU) algorithm would work? This might seem like a stupid idea until you consider that the most recently used pages are the ones that will still be in the disk cache, and so will be the least expensive to reload. Your simulation wouldn't show any gains because it doesn't take into account the relative costs of loading data which is in the disk's cache vs not in the disks's cache.<BR/><BR/>Nice work.Anonymousnoreply@blogger.com