On Tue, 2007-10-23 at 15:00 -0700, alan wrote: > On Tue, 23 Oct 2007, Mark Haney wrote: > > >> Now you really have me confused. "a system without directories"? From > >> where are you getting this information? I don't recall anyone saying > >> anything such as this. > >> > >> > > > > I, too, am interested in this. I know ext4 is the next version of the ext > > filesystem, but everything else mentioned by you has to my knowledge NOT been > > hinted at on any of the lists I'm on. Facts, man, give us facts. > > The only thing I have seen about a filesystem that does not use a > directory structure is blue sky handwaving from Microsoft. (Also known as > a "database filesystem" in some accounts.) How it would work is unclear. > How you would find anything is even more unclear. (And how such file > systems would handle millions of files is pretty unclear.) > > No one has implemented anything like this in any real usable fashion that > I know of. It would take an incredible amount of work to get Unix/Linux > programs to use it correctly. (Let alone secure it.) There are several commercial storage systems that use filesystems like that. Caringo's CAstor is one, Acinion is another. They use UUIDs to store the files (aka "objects" to them) and use a subset of HTTP 1.1 to do the I/O, as opposed to things such as NFS or CIFS. Yes, it's a little counterintuitive to those who are used to normal hierarchical systems such as we are, but it isn't blue-sky or FUD and it does work. Storage blocks are storage blocks. A filesystem is just ONE way of presenting the storage to the applications. Remember when we had to use CHS (cylinder, head, sector) addressing rather than LBA (logical block addressing)? ---------------------------------------------------------------------- - Rick Stevens, Principal Engineer rstevens@xxxxxxxxxxxx - - CDN Systems, Internap, Inc. http://www.internap.com - - - - To err is human, to moo bovine. - ----------------------------------------------------------------------
