On Thu, Mar 13, 2003 at 10:53:07AM +0100, Ralf Hildebrandt wrote: > Excellent. We encountered extreme performance degradation on our 1,2 > TB backupserver with ext3. Must have been due to the allocator. > With XFS, performance is much better. The allocator will probably help, but some of the extra overhead of ext3 is due to the fact that we're using indirect blocks. For very large files, the use of larger block sizes (i.e., 8k and 16k blocksizes on an ia64, and future work to support larger block sizes on systems with 4k page sizes) will help, but in the long run we may need to think about supporting an extent-based logical->physical block mapping. - Ted _______________________________________________ Ext3-users@redhat.com https://listman.redhat.com/mailman/listinfo/ext3-users
