Messages in this thread | | | Date | Wed, 12 May 1999 23:07:15 -0700 (PDT) | From | Erik Walthinsen <> | Subject | Re: Linux & NFS caching: reducing TCO |
| |
On Wed, 12 May 1999, Colin Hirsch wrote:
> In such a case it might be necessary to let the server be able to send a > kind of invalidate message to the clients if something changes (I > thought of keeping data in cache for tens of minutes). This whole setup > obviously relies on only few changes and only makes sense for e.g. /usr > but _not_ /home.
The problem with this is that NFS is generaly considered stateless. Sure, you can do NFS over TCP (though IIRC it's rather minimal in current Linux kernels), but that's not really enough state AFAIK.
In your case, I'd suggest using Coda. I was talking today with one of the guys that worked on Coda a while back about its problems and disadvantages, but your specific case just happens to be precisely what Coda was designed for...
I assume all the client machines still have disks from when they weren't diskless? What you do is use nfsroot to boot them, start up Coda and link it to the Coda server, use the local disk for caching/hoarding, and symlink /usr and /home to /coda/usr and /coda/home. All should be well, network traffic drops to very little because the clients will cache most or all of /usr on their local disks, and everything stays nicely in sync.
I've been thinking about doing a similar thing with a robot I'm building, but there I have the problem of needing to do the entire thing without a spinning disk, and without NFS entirely once it's booted. A lot messier than what you'd be up against, trust me... ;-)
Coda's home page is http://www.coda.cs.cmu.edu/.
TTYL, Omega
Erik Walthinsen <omega@cse.ogi.edu> - Staff Programmer @ OGI Quasar project - http://www.cse.ogi.edu/DISC/projects/quasar/ Video4Linux Two drivers and stuff - http://www.cse.ogi.edu/~omega/v4l2/ __ / \ SEUL: Simple End-User Linux - http://www.seul.org/ | | M E G A Helping Linux become THE choice _\ /_ for the home or office user
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |