SSSWG - February 1994 Minutes
IEEE SSSWG
February 8, 1994
IEEE SSSWG Attendance February 1994
NAME COMPANY
_______________________________________ _______________________________________
Betty Jo Armstead SVERDRUP Tech, Inc.
Bob Baird Hewlett-Packard
Peter R. Berard Battelle, Pacific Northwest Labs.
Ramesh Bodapati Hewlett-Packard
Carolyn Buford Computer Sciences Corporation
Walter A. Burkhard University of California, San Diego
Bob Coyne IBM, Federal
Jeff Deutsch IBM - Federal Systems Company
David J. Donald Metrum Inc.
Richard Garrison Martin Marietta Corporation
D. Creig Humes NASA Langley Research Center
David Isaac MITRE
Merritt E. Jones IBM / FSC
Steve Louis National Energy Res. Supercomputer Ctr.
Dennis Luck Department of Defense
Greg Nuss Cray Research, Inc.
Jon W. Peake IBM Storage Systems Division
Larry Pelletier NCR Microelectronics Products Division
Rich Ruef IBM/FSC
David E. Skinner Storage Technology Corporation
Jodie Smith Storage Technology Corp.
Linda L. Tefend E-Systems, Inc.
Richard Wrenn Digital Equipment Corporation
The topic of discussion was the relationship of IEEE SSSWG and
POSIX, particularly with the "Removable Media Study Group". One
possibility is to hold joint meetings with POSIX. They meet four
times a year and the registration fee is $300-$350 per meeting.
Another possibility is to meet at the same hotels as POSIX and
take advantage of their room rates, but arrange for our own
meeting rooms and stay with the lower registration fee. This
would mean meeting in hotels and would be more costly than
meeting at member sites. Bob Coyne agreed to investigate the
cost and report back later. The next POSIX meeting is at Incline
Village near Lake Tahoe.
The June (7-10) meeting will be in the TRI City area of
Washington sponsored by PNL. The members have agreed to have
lunch at the hotel ($10. per day) in exchange for lower rates on
meeting rooms. PNL will pick up the rest of the cost for the
meeting rooms. It was agreed to raise the registration fee to
cover the lunches to $130. The airport is at Pasco Washington
and the best access Delta through Salt Lake City. Also United
flies to Pasco through Seattle, however Peter did not recommend
the small plane over the mountains.
Bob Baird is investigating holding the August meeting at Asilomar
(near Monterey) Dave Skinner also offered to hold the August
meeting at Storage Tech..
Dave Isaac was ask to investigate holding the October Meeting at
Mitre to coincide to Goddard Mass Store Symposium.
Either Dave Skinner or Bob Baird will represent the SSSWG at the
IEE England Conference on April (4,5,6).
The 13 IEEE Mass Store Symposium will be in Annecy France in
June. This will be mostly a repeat of the papers given at the
12th Mass Storage Symposium held in Monterey. There will be a
SSSWG meeting with the Europeans on Friday afternoon and
Saturday. Bob Coyne does not particularly recommend that people
go since it will be primarily a repeat. Bob will be presenting a
a MSRM paper.
There will be a workshop on metadata late in the year sponsored
by Goddard and Fort Meade. Also a metadata strawman will be
presented some time this summer sponsored by Oak Ridge. One of
the reasons for these workshops is to attempt to establish some
common terminology. Currently there are as many languages
describing metadata as there are groups wanting to make use of
metadata. Examples are mass storage, imaging, and scientists
wanting keywords for searching data. These workshops are also
sponsored by the IEEE.
Bob Baird gave a report on a meeting with X/Open on a back up and
restore API. The API includes the ability to do a backup within
a data base. Basically it is a client server backup. They will
strip away a lot of the backup/restore standard. There was a
meeting with X/Open, Bob Baird and Rich Ruef. Bob will exchange
documents with them and try to come up with commonality with
SSSWG. This document specifies incremental backups and would
allow the client to incrementally backup a single large file.
Bob suspects the document will be massaged and come back much
changed. Bob was concerned because the terminology was different
from our terminology. Rich Wrenn believes this would be a peer to
the VSS and a client of the PVL.
Dave Skinner gave a short report on meetings with Posix. Dave,
Rich and Mike met with POSIX in Irvine. The person from Fermi Lab
is really pushing the standard. They discussed four work areas:
Store files on tape as a mountable files system.
Data sets on tape - how to create semantics and API
Defining Device semantics - Posix does not defines these at
present.
Overlap with the PVL - a different set of calls for the
PVL. Want a PVL with a different look and feel.
Rich Wrenn says there was a view that one does not mount a tape
but mounts a file system or data set. To Rich this is very
different from SSSWG PVL and is limited by the Posix definition
of a files system.
Dave Skinner and Rich believe there is an 80% overlap with the
SSSWG PVL. The people from POSIX do not like the interface to
the SSSWG PVL. They really want the ANSI Magnetic tape
sub-system. Another revelation is that they do not view the
system as layered. Rich Wrenn does not believe that Posix open
can position to a file. They talked about changing open
semantics to allow an open to a file. They have a reflector that
discusses all of these issues. By the next meeting they expect
to have the ground work for an API. They have a lot of concern
about vapor ware and do not want a standard that depends on an
API for which there is no software.
Rich Wrenn believes that if Posix puts together a mount service
and we put together a PVL one of the two will fail. To prevent
this Rich feels that SSSWG must meet in the same location as
POSIX, at least part of the time.
Thursday Afternoon
Bob Coyne reported on the possibility of holding the next meeting
at Lake Tahoe with POSIX. One possibility is to have the five
subcommittee chairs rent a room with a parlor at a cost of $150
per day as opposed to the $93 per day for the rest of the
working group. The difference in room rates would be assumed by
SSSWG. These five parlors would be used for the meeting rooms.
Additional costs would include the audio visual (about $1000).
the cost of the group dinner and snacks. The total cost will be
about $3500 out of pocket. Bob will try to negotiate a better
deal. This will mean people need to reserve rooms for Monday
through Thursday night at the designated hotel.
There was some discussion of holding the October meeting in
either New Orleans or Vancouver/Seattle (POSIX meeting) rather
than Washington D.C. at Mitre. No decisions were made.
Ramesh Bodapati discussed the routing problems in the Mass
Storage Reference Model. There is a need to have some knowledge
of the network topology. The optimum route between two movers
needs to be determined. Is there a need for a module like a
router module which knows where nodes are and the rates between
nodes? The router could then answer quires about best routes.
Actually here are two questions that need to be answered.
Is there a path between the two movers?
Can a class of service be guaranteed?
Rich Wrenn says the PVL must be able to ask two question can you
talk to mover A and what quality of service can you provide?
Dave Isaac said there could be a value added service to provide
the best path.
Bruce Haddon suggested that the Storage System must not
understand the routing across the network. The MSS can learn
about the network heuristically.
Bob Coyne suggested the system administrator could actually set
this up along with quality of service.
Ramesh has learned that the IP folks are tackling the problem of
providing a level of service. They will not provide the
connectivity.
Dave Isaac has pointed out that EOS is going to require
ISOCHRONOS service.
Bob Coyne suggested that the problem be given to the environment
group. Maybe this all belongs with the SSM - and not the storage
server.
Bob Baird wants a way to determine level of service and does not
care who provides it.
VSS - Bob Baird - status report.
Bob has two documents, his VSS document and the HPSS document
which will be merged by the next meeting. Bob is going to do the
merge. The document will still have issues to be resolved. Dave
Skinner would like an overview that he could review with out
reading the whole document and tell if there are points of
conflict between VSS and PVR. HP is attempting to write a
Storage Server using the Version 5 Model.
Dave Isaac believes that there needs to be some consistency of
whether there are transactions at the API interface. If so each
part of the MSS should use the same semantics. He suggested the
PVR subcommittee produce a sample that can be passed out to the
other subcommittees. Bob Baird believe the PVR is a subset of
what is required for the other servers. Where there is
commonality the other servers should use the same documentation.
Dave Isaac will take a shot at providing a style page to be used.
Bruce Haddon brought up the following issues:
- Question of negotiation of what services are being provided.
- All components should take the same approach.
Rich Wrenn says the client can know from the objects present
whether the function is available or not. If an object is
implemented, commands defined for that object must be
implemented. There will also be a version number available for
evolution of the MSS. Rich Wrenn has a problem of making the
object request broker a requirement of the MSS.
Bruce tried to bring the question back to "isn't this a MSS wide
problem and shouldn't there be a single solution". Should this
be given back to the people working on environment. Issues in
negotiation should provide more than just a yes or no answer.
For example a request for n bytes of storage should return back
more information than yes or no. Perhaps the amount of storage
available and/or what it will cost.
Bob Coyne's final comment was Andy Hanushevsky this is your life.
Steve Lewis presented a new drawing showing the relationship of
the SSM. It shows that, if the server does not provide the
managed object that the SSM might require, there can be a value
added P1244.5 part of the SSM that will attempt to use the
servers managed objects to construct the required objects. If
the server does not keep track of things needed by the SSM they
will come back and request that the added information be kept.
For example the PVR might simply provide an event with the
necessary information and the SSM would collect the information
required. The SSM will give the server requirements.
The PVR has one interface and two broad classes of service. The
class is separated by access control on services. Each
directive is classified as being management, client or both. The
directive are controlled by access control lists. Posix .6
(access control lists) can be used in the standards.
Bob Baird was going to provide an event for security and get back
a yes or a no. He feels that access control lists is being too
specific. He suggests that an event be defined for the MSS that
can be used by all servers.
======================================================================
Betty Jo Armstead xxbja@lerc.nasa.gov (216)-433-5086
The future belongs to those who believe in the beauty of their dreams
and work at making them come true.
SSSWG Home