(Illustration by Gaich Muramatsu)
I think what he's asking is whether Coda can export local files the way NFS servers can. One of the problems this would raise would be that it breaks Coda's consistency model (where the servers are assumed to be the true home of data, and the only changes to data on the server come from coda clients through the normal protocol). This of course assumes that one can update the local files. Here's why I'd like such a thing. We just bought a big netappliance file server that speaks nfs. Our central facilities will only backup data stored on the netappliance (don't ask why:( Ideally, I'd like the Coda server to access these files via NFS, e.g. the coda files are really nfs files. However, this introduces wierd failure modes. Alternatively, I could keep coda files in the local file system and copy them into NFS when they get updated (as a side effect of store). This would mean coda clients get the right semantics, yet I could "export" coda files to other nfs clients via our NetApp box. The only thing this really requires (I think) is a way of mapping the coda server files to names in the NFS world. One way is to replicate the coda name space seen by the clients in the NFS world. This would require venus-like functionality on the server. Alternatively, if you had enough disk space you could have a venus do the copies to NFS as a side effect of caching, but that would be substantially ugly imho. Anyone working on something like this? just musing... david. At 01:08 PM 3/24/98 -0500, Scott Smyth wrote: >Henry; > >The implementation explanation was not clear because I am not >sure if it would do what I was thinking. I was thinking that I would >use >Coda to cache read-only media (CD-ROMs) on the client side as Coda >does from the media server (Coda server and SCM). However, the server >and client would be the same. Thus, a media server would have dynamic >caching if you will and the user-level security of Coda going for it. >However, >in this context, I think Coda may be a little "heavy" for such an >implementation. >It just seemed like a good possibility to use Coda schemes tailored to a >different >use where there could be a heterogeneous network. > >Scott > >> -----Original Message----- >> From: Henry Pierce >> Sent: Tuesday, March 24, 1998 12:35 PM >> To: 'coda-discuss_at_coda.cs.cmu.edu' >> Cc: 'linux-coda_at_coda.cs.cmu.edu' >> Subject: Re: Using Coda for caching on local machine rather than >> over network? ?? >> >> >> I am not clear on precisely the question you have in mind about Coda, >> CDROM's and Ext2fs. All of these services are autonomous. If you >> had a clearer example of what you would like to try, I would be happy >> to be of further assistance. >> >> As for running a Coda Client and a Coda Server on the same machine, >> there are are only two issues you need to be aware of: >> >> + Virtual Memory >> + Disk Space >> >> As mentioned in our documentation, running a Coda Server, Coda Client >> and X-Window on the same machine requires somewhere around 64MB of VM >> for a moderately configured Coda-Server and Coda-Client. So, if you >> only have 32MB of VM, for example, , I wouldn't recomment it. >> >> As for disk space, you will need: >> >> Coda-Client-Cache-Space + Coda-Server-RVM-Space + Coda-Data-Storage >> Space. >> >> These can add up to quite a lot on one machine very quickly. >> >> regards, >> henry m. pierce >> >> >> ------------------------ >> Henry M. Pierce >> Research Programmer >> Department of Computer Science >> Carnegie Mellon University >> 5000 Forbes Avenue >> Pittsbugh, PA >> >> >> hmpierce_at_cs.cmu.edu hmp_at_infomagic.com >> >> On Tue, 24 Mar 1998, Scott Smyth wrote: >> >> > Date: Tue, 24 Mar 1998 11:10:08 -0500 >> > From: Scott Smyth <ssmyth_at_realminfo.com> >> > Reply-To: coda-discuss_at_TELEMANN.CODA.CS.CMU.EDU >> > To: "'linux-coda_at_coda.cs.cmu.edu'" >> <linux-coda_at_TELEMANN.CODA.CS.CMU.EDU>, >> > "'coda-discuss_at_coda.cs.cmu.edu'" >> <coda-discuss_at_TELEMANN.CODA.CS.CMU.EDU> >> > Subject: Using Coda for caching on local machine rather than over >> network? ?? >> > Resent-Date: Tue, 24 Mar 1998 11:14:12 -0500 >> > Resent-From: coda-discuss_at_TELEMANN.CODA.CS.CMU.EDU >> > >> > In the context of a read-only system being cached...say >> > a CD-ROM to a linux ext2 partition: could Coda be implemented >> > to act as a cache fs on the same machine (i.e., server=client) >> > rather than over a network while maintaining Coda features locally >> > just as a would over a network? Are there any problems that >> > would arise between the client, the server, and the SCM being >> > the same machine? >> > >> > Thanks, >> > Scott >> > > > >Received on 1998-03-24 14:29:29