| PLEX86 | ||
|
fragmenting file systemsYou could I suppose take a 'brute force' approach also. Clumsy but should work on most any file system: start with blank FS write it full of say 10MB numbered files. remove every other one ( like all the odd numbers) then fill the remaining space with 20MB files. remove the remaining 10MB files What's left is (roughly) a whole bunch of fragmented 20MB files. Fill the remaining space with 40MB files and you'll have then a whole bunch of fragmented 20 and 40MB files, all roughly in 10MB chunks. exact details will vary of course but change the numbers and procedure and you could probably create about any sort of fragmentation you like. I don't know how "% fragmented" would be defined but it would be fairly deterministic. Also depending upon the actual FS details the fragmentation could be much greater then this simple-minded analysis indidcates. But buttuming ( I know I know..) that the FS tries to write the file in a contiguous manner, the fragmentation should be pretty predictable. Worst case this should be lower bound- it could only get more fragmented. multisecond hang in getdents Linux 2.4, x86, ext3fs I'm seeing a strange performance problem on one of our systems. When running a find on one filesystem, there are long hangs (up... Throw in some random values and you could get a more realistic result Unanswered question is of course: Why? Perhaps some sort of experiment? Stan -- Stan Bischof ("stan" at the below domain) www.worldbadminton.com
|
||||
multisecond hang in getdents Linux 2.4, x86, ext3fs Linux groups from Newsgroups The #1 Usenet Provider on the Internet
|
||||