Ah, doesn't NFSv3 not support concurrent reads? Or concurrent reads while a
file is open for writing or something like that? Concurrent writes would be
useful, but is not necessarily required (we could hack around it). Idealy,
we'd like to have a job continue to read off the version of a file it
started reading on, such that when the file already opened for reading is
opened for writing, the file is duplicated, the duplicate writen to (now the
new version), then deleted once the initial job closes the file for reading.
This could result in many versions of the same file, but only as long as
they are open for reading. It would be nice if I could open a file for a
"sustained read" or some such, so only big jobs would receive the special
treatment of getting their own temporary version of a file.
>NFSv4 for Linux is about three months out. a beta quality
>
> > client will be available soon, but the server is farther
>
> > out.
>
>
>
> >
>what are your performance requirements, and why do you
>
> > think NFSv3 won't meet them?
>
>
> > I'm setting up a distributed/clustering filesystem for a
>
> > > research group and
>
> > > would like some info/advise.
>
> > >
>
>
> > > NFSv4 looks pretty sweet, but is it ready for beta
>
> > > deployment? We'll need to
>
> > > have everything up and stable
>fairly soon, but so far it
> > > seems like
>NFSv4 is
>
> > > the only open source FS out there that will satisfy our
>
> > > requirements, of
>
> > > which the main ones are
>partial-file caching and support for
> > > large
>(100GB+)
>
> > > filesystems.
>
> > >
>
> > > Intermezzo forces whole
>file caching, which will not do.
> > >
>OpenGFS I believe
>
> > > is still in alpha stage. OpenAFS only allows for very
>small
> > > filesystems.
>
> > > EVMS looks
>promising, but the clustering FS part is still
>
> > > alpha. I think
>
> > > NFSv3 has poor performance, although it would be
>excellent if
> > > it would work.
>
> > >
>
>
> > > I'd greatly appreciate any info on distributed/clustering
>
> > > filesystems,
>
> > > especially NFSv4 or stable
>alternatives.
> > > Thanks much,
> > William Knop
_________________________________________________________________
Add photos to your messages with MSN 8. Get 2 months FREE*.
http://join.msn.com/?page=features/featuredemail
-------------------------------------------------------
This sf.net email is sponsored by: To learn the basics of securing
your web site with SSL, click here to get a FREE TRIAL of a Thawte
Server Certificate: http://www.gothawte.com/rd524.html
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
>>>>> " " == William Knop <[email protected]> writes:
> Ah, doesn't NFSv3 not support concurrent reads? Or concurrent
> reads while a file is open for writing or something like that?
Something like that: it doesn't support calling back clients in order
to notify them of changes that occur on the server because the model
is designed to be stateless and thus to scale for large numbers of
clients.
Although NFSv4 adds the concept of state (and hence file locking) to
the protocol, it does not differ from NFSv3 in this respect.
You probably want to be looking at dedicated clustered filesystems
such as OpenGFS, for instance.
Cheers,
Trond
-------------------------------------------------------
This sf.net email is sponsored by: To learn the basics of securing
your web site with SSL, click here to get a FREE TRIAL of a Thawte
Server Certificate: http://www.gothawte.com/rd524.html
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
I'd like to use OpenGFS, but from what I've heard it's very not useable. If
someone would tell me otherwise, I'd be happy to give it a shot.
>From: Trond Myklebust <[email protected]>
>To: "William Knop" <[email protected]>
>CC: [email protected]
>Subject: Re: [NFS] Re: status
>Date: 18 Nov 2002 19:40:09 +0100
>MIME-Version: 1.0
>Received: from mons.uio.no ([129.240.130.14]) by mc6-f6.law1.hotmail.com
>with Microsoft SMTPSVC(5.0.2195.5600); Mon, 18 Nov 2002 10:40:12 -0800
>Received: from charged.uio.no ([129.240.86.49])by mons.uio.no with esmtp
>(Exim 2.12 #7)id 18Dqoc-0007XR-00; Mon, 18 Nov 2002 19:40:10 +0100
>Received: from trondmy by charged.uio.no with local (Exim 2.12 #1)id
>18Dqob-00069i-00; Mon, 18 Nov 2002 19:40:09 +0100
>References: <[email protected]>
>In-Reply-To: <[email protected]>
>Message-ID: <[email protected]>
>Lines: 17
>User-Agent: Gnus/5.0808 (Gnus v5.8.8) XEmacs/21.4 (Common Lisp)
>Sender: Trond Myklebust <[email protected]>
>Return-Path: [email protected]
>X-OriginalArrivalTime: 18 Nov 2002 18:40:13.0155 (UTC)
>FILETIME=[EE27CB30:01C28F31]
>
> >>>>> " " == William Knop <[email protected]> writes:
>
> > Ah, doesn't NFSv3 not support concurrent reads? Or concurrent
> > reads while a file is open for writing or something like that?
>
>Something like that: it doesn't support calling back clients in order
>to notify them of changes that occur on the server because the model
>is designed to be stateless and thus to scale for large numbers of
>clients.
>Although NFSv4 adds the concept of state (and hence file locking) to
>the protocol, it does not differ from NFSv3 in this respect.
>
>You probably want to be looking at dedicated clustered filesystems
>such as OpenGFS, for instance.
>
>Cheers,
> Trond
_________________________________________________________________
STOP MORE SPAM with the new MSN 8 and get 2 months FREE*
http://join.msn.com/?page=features/junkmail
-------------------------------------------------------
This sf.net email is sponsored by: To learn the basics of securing
your web site with SSL, click here to get a FREE TRIAL of a Thawte
Server Certificate: http://www.gothawte.com/rd524.html
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs