Messages in this thread | | | Subject | Re: EXT2 and contiguous files | Date | Fri, 09 Jan 1998 17:50:28 -0800 | From | Craig Milo Rogers <> |
| |
> Of course, for the ultimate in performance gain, you would have >to change the on-disk format, and the way to do this compatibly would be >to use an ext2 inode flag to designate whether an inode is using the old >format or the new format. As you noted, for sparse files, for obvious >reasons the old format is superior.
Sigh. For the ultimate performance gain, for small files (with size less than some threshold), surely it is best to store the data of the file directly in the inode. In some ways, this parallels the decision on whether to store a symbolic link in the inode or in a separate data block.
Per a different discussion thread on this mailing list, when files are singly linked into directories which are a certain size relative to system memory, and for some set of application file access statistics, surely the best thing to do is to store the inode into the directory. In other cases, of course, it's a real loser.
Thus, there probably exist circumstances in which the best thing to do is to store the inode and the file data both directly in the directory. I'd hate to have to fix one that got clobbered, though.
Craig Milo Rogers
| |