On Wed, 26 Nov 2003 14:24, Bernd Eckenfels <[EMAIL PROTECTED]> wrote: > > I am talking about any file system. When only reading from a file system > > there should not be any performance difference when comparing a RO mount > > vs a NOATIME mount. If there is a difference then it's a bug in the file > > system. > > I guess the thread was about non-journalling filesystems beeing faster, and > less of a risk if used ro.
Even for a non-journalling file system there should be no risk. If a file system is mounted and never written to then only a single disk block should change, the one with the dirty bit indicating that an fsck might be needed on a reboot. If that block is corrupted then you may need to use the backup superblock in the worst-case, but that would require a crash while mounting the file system. > > The difference it makes is that reading from the disk will never cause > > disk writes. If you access large numbers of files or if you have IO > > hardware that has a bottleneck of write bandwidth (EG a typical mail > > server) then NOATIME makes a significant difference. > > News Servers are even worth. And full-filesystem scans and some backup > tools make the a-time less usefull anyway. There should be a way of reading a file without changing the ATIME that backup programs can use. -- http://www.coker.com.au/selinux/ My NSA Security Enhanced Linux packages http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark http://www.coker.com.au/postal/ Postal SMTP/POP benchmark http://www.coker.com.au/~russell/ My home page
