Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!elroy.jpl.nasa.gov!turnkey!orchard.la.locus.com!fafnir.la.locus.com!werth.la.locus.com!richard From: richard@locus.com (Richard M. Mathews) Newsgroups: comp.unix.aix Subject: Re: But what about less INODES (Re: Need more inodes) Message-ID: Date: 23 Apr 91 01:24:22 GMT References: <1991Apr16.151844.24793@edm.uucp> <1408@cameron.egr.duke.edu> Organization: Locus Computing Corporation, Los Angeles, California Lines: 24 jpe@egr.duke.edu (John P. Eisenmenger) writes: >I noticed something yesterday that I'm not happy about in the least. When >I ran "du" (while setting up "spacegripe") I noticed that all of the file >block sizes were multiples of 4. In other words, it looks like IBM decided >to not have any fragmented blocks. Being that my machine will be the target >for many, many small files (spice input files, etc.), this is a very unat- >tractive "feature" although I can see how it would speed up the filesystem. You don't mention what platform you are on (please, everyone, platform and release information is very helpful). I don't know what AIX on the 6000s does. AIX/370 and AIX PS/2 lack a real fragmentation scheme, but they do have a way of storing files of less than 384 bytes in the space normally used for inodes. The "du" and "ls" commands will show these as using up 0 blocks. Unfortunately, the 384 byte limit is small enough that not much more than small directories (and most importantly, hidden directories) can take advantage of this. The problem in AIX/370 and AIX PS/2 has been with combining fragmentation with the shadow block mechanism used to atomically commit or abort file system changes. Richard M. Mathews D efend richard@locus.com E stonian-Latvian-Lithuanian lcc!richard@seas.ucla.edu I ndependence ...!{uunet|ucla-se|turnkey}!lcc!richard