Coda File System

Re: FW: Using Coda for caching on local machine rather than over netw ork? ??

From: <braam_at_cs.cmu.edu>
Date: Tue, 31 Mar 1998 06:07:30 -0500
Andreas Jellinghaus writes:
 > IIRC coda can only export files stored in it's own rvm partitions.
 > that's necessary for coherenz, versioning, journaling, logging,
 > additional management like access control lists, etc.
 > correct ?

Almost correct.  RVM is used for Metadata -- acl's, vnode's, version
vectors (for server replication), directory contents. This is managed
through a transaction system, and we feel it would be very hard to do
it without transactions, since during resolution/repair/reintegration
many changes are being made, but one may have to back out (e.g. if a
server crashes, or if an unexpected inconsistency arises).   

We use RVM for transaction support -- as in most transaction systems,
it has a "DATA" part (the metadata) and a LOG part, containing the
modifications until they are integrated into DATA.  We could
potentially use a transaction based database system instead.

The file data is stored in Coda "partitions".  Files are references by 
a number and this number can be interpreted in various ways: ideally
one would access the file by inode number directly, currently we use a 
tree system to access files. 

 > 
 > and there are better ways to do that :
 > benfs, user fs, onion fs or how they all are named - filesystems either
 > to cache a cdrom, or to make it possible to change it.

It would probably be possible fairly easily to add read-only serving
of CD Rom's to Coda fairly easily.

- Peter -


 > 
 > andreas
Received on 1998-03-31 06:10:52