[OpenAFS] key issues for OpenAFS future?

[Neulinger, Nathan]

OpenAFS works fine with journalled filesystems. The only incompatability
is that you cannot use reiserfs as your cache partition with 2.4.x+
kernels since reiserfs does not support retrieving a file from disk from
just the inode number, unlike most other unix filesystems.=20

Ext3 works just fine as a cache partition, and ANY of the filesystems
works for server vice partitions.=20

Without _MAJOR_ protocol changes, afs will never support byte range
locks. I looked into various approached on this a while back, and the
problem is that the cache manager does not guarantee anything other than
whole file updates, so even if the protocol could support them over the
network, it wouldn't be safe without changing how the cache manager
behaved.

I believe that when we last discussed this, about the best that can be
hoped for here is a client-side/in-libafs lock manager that would manage
and track byte range locks on a single client, but upgraded those locks
to a single whole-file lock on the server. i.e. multiple apps on a
client could use byte range locks, but it could not cross a client. This
may or may not be sufficient for most purposes.

I believe the 2GB filesize issue is being worked on at least to some
degree, but not anything really targetted.

As for the "not going to be a serious contender"... I would tend to say
that OpenAFS is the _ONLY_ serious contender for a real distributed file
system in the market. DFS (the real DFS) is way overkill and completely
unavailable to most users. If they think Coda or Intermezzo are, well,
anyone who would answer that in that way will never know the difference.

-- Nathan

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-4841
Computing Services                       Fax: (573) 341-4216


> -----Original Message-----
> From: Paul Blackburn [mailto:mpb@est.ibm.com]=20
> Sent: Tuesday, November 05, 2002 1:35 PM
> To: openafs-info@openafs.org
> Subject: [OpenAFS] key issues for OpenAFS future?
>=20
>=20
> Hello,
>=20
> I have seen a comment that:
>=20
> >"OpenAFS is interesting in certain situations, but until it=20
> has support for
> >things like files over 2 GB, byte range locks, and journaled=20
> filesystems
> >it is not going to be a serious contender in the filesystem market."
> >
> For the way we use AFS, none of these are an issue.
> I can see that the 2GB filesize could be a consideration for=20
> some sites.
>=20
> For me a more important issue would be getting AFS kernel modules
> into the Linux kernel source tree because this would enable porting
> to many distros and machine architectures "at one go".
>=20
> What do other people think?
> What capabilities are considered important for the OpenAFS future?
> --
> cheers
> paul                          http://acm.org/~mpb
>=20
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info
>=20