Coda File System

Re: some thoughts and questions

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Sun, 9 Mar 2003 14:43:56 -0500
On Sun, Mar 09, 2003 at 12:17:26PM +0100, Ivan Popov wrote:
> So my question is - how much would have to be rewritten to be able to use
> some other database, allowing the same persistency and resilience,
> hopefully comparable performance, but not dependent on mapping the whole
> data in memory at once.

Not sure how much would have to be rewritten, but a lot (most) of the
server code isn't really using RVM all that much. Possibly because rvm
was added on at a later point and most of the real smarts and challenges
were in the Coda clients.

As a result, most accesses to vnodes and volumes use functions like
GetVnode/PutVnode, which return a 'cached' VM copy of the RVM data. When
the VM copy is modified, it is written back to RVM during the PutVnode
operation. Same holds for directory data.

As far as venus is concerned, RVM is very much integrated everywhere and
it would be a very significant effort to replace with anything else.

Jan
Received on 2003-03-09 14:45:43