On Feb 20, 2008 5:48 AM, Douglas A. Tutty <[EMAIL PROTECTED]> wrote: > While this kind of setup is well beyond my pay-grade, looking just at > the issue of, in effect, using threads to share a cache to avoid hitting > the disk, I wonder why using a memory filesystem as the common cache > wouldn't work. No threads, shared data via the filesystem but that > filesystem is in memory and quite fast.
Because the data structure actually used by your program is rarely bit for bit identical with the on disk representation of that same data. Memory filesystems impose all sorts of overhead like inodes, names, directories, modification times, sizes, owner id, group id, access time, permissions, superblocks, backup superblocks.
