[PVFS-developers] sc2003 BoF wrapup

[Robert Latham]

Thanks to everyone who came to the PVFS BoF session during SC2003.
The official attendance count was 90 -- and that was with some stiff
competition from several other interesting BoF sessions.

For those of you who may have missed it, we talked a bit about PVFS1,
then spent most of the time discussing PVFS2.   You can see the
presentation slides here:

PVFS1 status:
http://www.mcs.anl.gov/~robl/pvfs2/sc2003-bof-pvfs1.pdf

And presenting PVFS2:
http://www.mcs.anl.gov/~robl/pvfs2/sc2003-bof-pvfs2.pdf

We got some good questions from the audience that (naturally) are not
captured in the above slides: 

. People were quite curious about PVFS2 performance.  We've spent most
  of our time at Argonne testing for correctness and scalability, but
  we have been able to hit about 7 GByte/sec aggregate bandwidth when
  all the caches are warmed up.  

. Evidently Sun has stopped work on their parallel file system.  There
  was at least one person interested in PVFS on Solaris.  

. People were curious about adding and removing servers to/from a PVFS2
  volume.  Right now, the hardest part is making sure all clients know
  about the change in servers.

. We talked a bit about optimizations we can make for read-heavy and
  read-only applications (and it turns out there are quite a few of
  those applications).  RobR went out on a limb and promised support
  for an "immutable" flag by SC2004 -- even if it meant doing a
  bunch of work a couple weeks before the show :> 

. There was a question about encrypting files on PVFS2, but we'll
  probably leave that up to applications or ambitious researchers

. We even had some words of praise.  One for providing "an alternative
  to GPFS", and several people happy that pvfs1 just works and works
  until they have some hardware problem (usually a disk failure). 

If you were there, and i didn't cover a point you thought was
important, speak up and fill in the details for the folks who couldn't
make it.

==rob

-- 
Rob Latham
Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
Argonne National Labs, IL USA                B29D F333 664A 4280 315B