Coda File System

Re: db and coda

From: Peter J. Braam <braam_at_cs.cmu.edu>
Date: Wed, 07 Apr 1999 11:18:19 -0400
Hi 

Coda does whole file caching - not very desirable for a large database.
What you need is write/write sharing of a single file, and with Coda
this isn't really a good idea, since

1. we fetch the entire file
2. files only get updated on the server when they are closed on the
client

This was a deliberate design choice that is not so compatible with
database use. 


- Peter -



Radha Krishna Pagadala wrote:
> 
> On Tue, 6 Apr 1999, Laszlo Vecsey wrote:
> 
> > As I understand things, one of the nice things about oracle is that the
> > servers can be set up with true redundency and fault tolerance. i.e.
> > multiple servers can fail and the database will still be accessible.
> >
> > Would using gdbm or db (sleepycat db, included in GNU glibc), on a
> > distributed coda filesystem layered on top of raid5 do the trick?
> >
> > I know coda efficiency isnt there yet, but can someone comment on how the
> > above solution would compare in the foreseable future..
> >
> > - lv
> 
> as far as i understand coda has not been designed for such use, the
> thesis of kistler specifically states that coda does not support
> distributed databases on top of it, coda works best in an environment
> where 'writes' are infrequent, and 'reads' are frequent and where
> concurrent updates are rare
> 
> conditions that i think will not be met with the intended application
> 
> -radha krishna
Received on 1999-04-07 11:19:58