Coda File System

Re: FAQ: Is Coda ready for use?

From: Joerg Sommer <joerg_at_alea.gnuu.de>
Date: Tue, 2 Dec 2003 10:32:05 +0000 (UTC)
begin  "Stephen J. Turnbull" <stephen_at_xemacs.org> wrote:
>>>>>> "Lionix" == Lionix  <lio_at_absium.com> writes:
>
> I don't see why.  Maybe multiple servers, multiple VSGs for
> performance or organizational reasons, but Coda scales well in that
> direction.

Nice to hear.

> RVM (where the directory structure is kept) can be disabled when
> initializing the client, but this means you need to keep all
> information in memory, and of course it's not persistent.  Increases

Is a memory cache possible? Today one GB RAM is almost normal and there
300MB for a cache are possible.

> startup costs and risk of mutation loss if there's a crash before
> saving.

Persistence doesn't matter. The machines running in 24/7 mode.

>     >> Does coda need a process in userspace to access the coda fs?
>
>    Lionix> If you mean a process started by the user the reponse is
>    Lionix> no.  Coda must be started by the root, init script stuff.
>
> Ie, "yes".  The coda client is quite complex; the kernel provides only
> enough services to implement to VFS interface, so that ordinary
> programs can use usual file APIs.  All real work is done by upcalls

Why? Is there a reason, why the venus daemon can't be in the kernel? i.e,
NFS or SMB doesn't needs a userspace daemon.

>  [Graph]

Very good explanation.

Joerg.
end.
-- 
Received on 2003-12-02 06:08:32
Binary file ./codalist-2003/5875.html matches