SSSWG - March 1996 Minutes
IEEE Storage System Standards Working Group Meeting
19th-21st March, 1996
Boulder Broker Inn, Boulder, Colorado.
Attendance:
Carl Braganza carl@acsc.com ACSC, Inc
Bob Baird bbaird@cup.hp.com Hewlett-Packard
Mark Carlson mac@RedCape.com Red Cape Software
Jack Cole cole@arl.mil
US Army Research Laboratory
Stuart Craig scraig@exabyte.com Exabyte Corporation
Ole Fischer ole_fischer@bbn.hp.com Hewlett-Packard
Bruce K. Haddon bkh@spot.colorado.edu Consultant
Dixon Hutchinson hutch@exabyte.com Exabyte Corporation
Merritt E. Jones merritt@mitre.org The MITRE
Corporation
John Merrill jhm@ncar.ucar.edu NCAR
Gary Mueller Gary_Mueller@stortek.com
Storage Technology Corporation
Dave Skinner Skinner@netcom.com Red Cape Software
Eric Stouffer stouffer@vnet.ibm.com IBM/Tucson
Mike Wenzel mw@core.rose.hp.com Hewlett-Packard
Joel Williams joel.williams@SES.ses-inc.com
Systems Engineering and Security
Rich Wrenn wrenn@cookie.enet.dec.com
Digital Equipment Corporation
Chair: Dave Skinner
Recording Secretary: Bruce K. Haddon
General Session, Tuesday, March 19, 1996, 8.25.
1. Opening Remarks, Introductions
==============================
Stuart Craig, Ole Fischer, and Mike Wenzel were welcomed to their
first attendance, and everybody around the table introduced
themselves.
1.1 Description of products or goals of attendees.
Exabyte: tape and library manufacturer, host of the current
meeting.
Red Cape Software: new start-up company, with a product in
the storage management area, portable to many platforms. Mark
Carlson spoke of being involved in the Desktop Management Task
Force, and that the DMTF was standardizing a RAID MIB.
HP: Is conducting research for fiber channel-based mass
storage, as well as being interested in storage management
DOD: High performance computing, shared resource centers
MITRE: is a "federated laboratory" interested, at the
moment, in image intelligence.
NCAR: using model for next generation system designated
MSS4, which is to be more distributed than existing system.
1.2 Free Discussion of New Technology
No items.
1.3 Announcements
Finances of the group are in good shape.
2. Agenda, Minutes
===============
2.1 Agenda
The agenda was accepted as published prior to the meeting,
subject to changes as needed as the meeting proceeds.
2.2 Minutes
Dave Skinner read the minutes from the January meeting.
## Resolved: that minutes be approved with one change, to item
7.4, which should have recorded that this discussion took place,
not that there was an action item that this discussion should
take place. The item in question had read as below, and is to be
deleted:
"Action Item: Addressing the issues raised by Mike Refi
and Denise Wright concerning the treatment of scratch
volumes by the PVR in tape stackers versus tape random
access magazines (Carl, Joel, Rich, Mike, Denise, Dixon)."
Moved: Rich Wrenn; 2nd: Joel Williams
Passed: unanimously
3 Matters arising from the minutes
================================
3.1 Robotics Level Interface
HP is sponsoring a standards proposal for a robotics level
interface. This would be a good time for this work to be
incorporated or coordinated with the work of the SSSWG and the
OSSI reference model
3.2 Length of the SSSWG meetings
## Resolved: that the SSSWG meetings be three days long rather
than 3 1/2 days.
Moved: Rich Wrenn; 2nd: Dixon Hutchinson
Passed: unanimously
3.3 The hiring of a technical writer
The plan to hire a technical writer to work on the PVR API
specification did not eventuate. Overall, with the work done by
Rich Wrenn and Joel Williams, the need to do this is no longer
felt. No further action will be taken at this time.
3.4 Availability of the standards drafts
It was suggested, again, that the drafts of the API standards be
available on the SSSWG web page, as is other documentation
produced by the group. Experience has shown that HTML is too
restrictive to express the richness of the documents that are
being generated. PostScript that can be ftp'ed was deemed
acceptable.
An electronic form of the MVR document does not seem to be
available. It is believed that Creig Humes has this document in
MS Word format or RTF format. Bruce claimed there was a document,
and Bob Baird was able to produce a copy.
Action Item: Dave Skinner will contact Creig Humes to find
out what there is in the way of a document, and make it available
somehow on the web page.
Action Item: Jack Cole has undertaken that the other
material from this meeting, including the API drafts (other than
MVR), and the minutes from this and past meetings will be moved
to the web page, or will be pointed to by the web page and be
available by FTP.
3.5 State of the API standards
At the last meeting, Rich Wrenn made presentations on the state
of the API standards, and in particular, discussed issues with
the standards, including the problem of scratch volumes and tape
stackers. It was the general opinion that the content of these
presentations was too important to be lost.
Action Item: Rich Wrenn will summarized some part of the
presentations, and make the result available for the web page
3.6 Schedule for, and Frequency of, Future Meetings
May: The next meeting will be held in May at the Aberdeen Proving
Ground, Maryland. Jack Cole, US Army Research Laboratory will be
the organizer. Notices will be posted to the reflector as soon as
details are known.
July: general approval was expressed that the July meeting
should be again in Colorado, and that a mountain site should be
investigated. Tuesday, 9th, to Thursday, 11th, was suggested as
the most opportune dates on which to hold this meeting.
Action Item: Rich Wrenn and David Skinner to investigate an
appropriate site.
September: suggested that this meeting be held
"concurrently" with the Goddard Mass Storage Symposium, which is
on Tuesday-Thursday, 17-18-19th September. This could be
accomplished by meeting on the Monday 16th, and Friday 20th.
Merritt explained that the Goddard MSS is *the* Mass Storage
Symposium for the year, it having been decided that the IEEE (US)
MSS, the IEEE (European) MSS, and the Goddard MSS should rotate
year-by-year.
Action Item: As proposed organizer, Joel Williams is to make
investigations of the practicality of doing this.
November: there was considerable discussion, as November is
a busy month, with election day, COMDEX, Super Computing
Conference, and Thanksgiving. The initial suggestion was that the
meeting could be held, perhaps, Wednesday-Friday, 6-7-8th, thus
avoiding election day on the Tuesday. The week beginning the 18th
has both COMDEX in Las Vegas, and Super Computing in Pittsburgh.
The alternative suggestion was made that the meeting could be
held in Las Vegas, Wednesday-Friday, 13-14-15th, with people then
being able to stay on over the weekend to join COMDEX the
following week.
## Resolved: that the option of combining as described above
with COMDEX is preferred.
Moved: from the Chair.
Passed: 6-0.
Action Item: Dave Skinner to investigate feasibility of this
arrangement.
Carl Braganza suggested that there be meetings only every
three months, i.e., four times a year. Discussion ensued, with
most people expressing a preference for the present "every two
months", as this helped support what was really minimal momentum.
Session adjourned at 9.55 a.m.
Session reconvened at 10.12 a.m.
General consensus was that six times a year continues to be
preferable. The discussion ended without any resolution to change
the current arrangement being proposed.
3.7 Review of Action Items from Last Meeting
(1) Refinement of the PVL API draft (Carl and Joel).
Done, and Joel will make a presentation on this subject later.
(2) Refinement of the PVR API draft (Dave Skinner and Rich
Wrenn).
Done, with Joel and Rich participating, and Joel will also
present on this subject at this meeting.
(3) Updating the WEB page with copies of the refined drafts.
This was discussed earlier. A new action item has already been
assigned to Jack Cole.
(4) Development of the scope of the ROBOTICS PAR (Hamid)
This action item will be determined by discussion scheduled to
take place Thursday.
(5) Arrangements for the March and May meetings (Dixon, Jack).
Dixon has arranged the March meeting. See discussion and action
items above with regard to future meetings.
(6) Determination of the status of other APIs such as MVR.
Not done. See new action item above, moving responsibility for
this to Dave Skinner from Jack Cole
4 Other Matters
=============
4.1 Presentation of SSSWG Work
Joel and Rich have proposals into the Goddard MSS for papers that
will discuss various aspects of the SSSWG work.
Joel is working on a "logical" tape format, that makes a
proposal for a directory on the tape, in the first partition, to
describe the contents, which are in the second partition. This
proposal is similar in spirit to schemes used by STK and IBM and
DEC (but all of those hide the directory structure within the
drive operation). There are issues related to metadata updates,
and space recovery. Exabyte supports partitionable tapes,
previously two partitions, now sixty-four.
Tape Format Study Group will be meeting in Chicago, April 1,
meeting as part of the AIIM (Association for Imaging and
Information Management). Contact Joel Williams for further
information.
There is an ISO group looking at the question of "digital
archives". In particular, there is a "reference model" that
includes and makes reference to the SSSWG reference model.
Information is available from their WEB page (contact Jack Cole
for details).
5 New Business
============
5.1 Model Revision
Dave Skinner spoke of his new company, Red Cape Software, perhaps
sponsoring a technical editor to work on a revision of the
reference model.
Merritt Jones spoke of the difficulty posed by the date of
the document, a date which is close to two years ago. Merritt
would like to see a minimal update, even if it is only to include
a list of known deficiencies, that would be available in 1 - 3
months.
Concern was expressed, particularly by Carl Braganza, Joel
Williams, and Rich Wrenn, that a "technical editor" would have
also to be a "model architect", and that there would have to be a
mechanism to ensuring input from the whole group to the architect
for the issues to be addressed, and setting direction for the
architect's resolution of those issues.
Action Item: Dave Skinner, by end of meeting will have a
resolution of the question of what to advertise about a revision,
to whom, and where.
5.2 Election of Chairman
Term of the current chairman, Dave Skinner, ends in August, 1996.
Dave undertook to finish his term. In order to have an election
at the July meeting, nominations need to be made at or by the May
meeting.
5.3 Review of Chair Positions for Sub-Committees.
Session adjourned at 11.48
Session reconvened at 13.24
A new Chair is needed for the MVR sub-committee
A new Chair is needed for the PVL sub-committee
PVR sub-committee needs someone who can focus more
attention, but Dave will continue at the current level if needed.
Rich Wrenn expressed the thought that PVR activities are not
being hindered by the current arrangement.
## Resolved: that the MVR Chair be Joel Williams
Moved: Merritt Jones; 2nd: Bob Baird; nomination accepted.
Passed: unanimously.
PVL Chair: issue tabled until May meeting
5.4 Review of API Work
Concern has been expressed about the complexity needed for APIs
and need for overviews to accompany APIs.
"Complexity" here refers to questions relating to:
How much detail is needed;
Whether there is should be multiple levels of compliance;
Whether there should be separate service and management
API's, and whether it is possible to publish one
without the other.
What is the definition of "location independence".
The need for overviews refers to some textual introduction
to the overall concept of the API. In particular, there needs to
be a list of assumptions about other components and the
environment, and also definitions of terms used in the
description of the functions and data structures API. There was
general consensus that each API composer should solve this
problem, with each API being internally complete and self-
contained. Merritt made the point that each standard, having its
own PAR, will be balloted independently.
5.5 Expiration Dates
Four of the PAR's will expire at the end of the year. By that
time, we will have to have a justification for an extension. The
PAR for the reference model has already expired. As the reference
model is not intended to be a standard, this may not be a
problem, but there is no mechanism in IEEE at present for
documents like the reference model.
6 API Development
===============
6.1 Review of Status
What are the current requirements for the model components where
there has been less activity, i.e., APIs such as MVR, MGT, etc.
What is the role of a MGT standard? Rich Wrenn summarized the
debate to date:
Are management interfaces the responsibilities of each dot
group
Is there a common framework (i.e., does MGT just choose this
framework)
Is the management interface specified within a framework by
profiles?
Should management methodologies be allowed to evolve?
What are really the underlying assumptions and philosophy?
Action Item: Dave Skinner proposes that at the next meeting
to offer a rationale for having a MGT group, and an outline for
an approach to the activities. Bob Baird will consider
contributing.
6.2 Review of PVR API Draft
Joel Williams made a presentation on the PVR API. The model of
the PVR implied in this API is based on the paper by Rich Wrenn
entitled "IEEE Storage System Standards Physical Volume
Repository Model (proposed)", which is available from the ftp
server.
Action Item: Joel will make the text of his presentation
available on the ftp server.
Discussion: The PVR model is specified in terms of certain
objects, plus "transfer mechanisms" and "controllers". These two
latter are not objects, but exists as methods of describing
certain relationships.
It became clear in some of the discussion that the model
underlying the API was itself not well-understood. Rich Wrenn
suggested that another discussion be scheduled to again visit the
PVR model, and concentrate at this time on the implementation of
the model in the API. It was agreed that this be done,
immediately following the API presentation. (Logically this
should be done in the other order, but the API presentation will
illuminate the questions needed to be answered in terms of the
discussion of the PVR model.)
Meeting adjourned at 17.01
A majority of the attendees planned on attending the tour of NCAR
arranged by John Merrill.
6.3 Review of PVL API draft
Joel Williams made a presentation on the PVL API in a breakout
session, held in parallel with the following review.
Action Item: Joel will make the text of his presentation
available on the ftp server
6.4 Review of the MVR API draft
Bruce Haddon reviewed the status of the MVR draft, and listed the
work still to be done: mainly in the area of event mechanisms,
MVR-MVR protocols, and security.
Breakout sessions continued during the afternoon
The session was adjourned at 17.00.
General session, Thursday, March 21, 1996, 8.48
7 Summary Sessions
================
7.1 PVR Breakout Session Report
Joel Williams reported on the PVR discussion from the breakout
session of yesterday. This discussion resulted in the following
list of things to do:
(1) Add "list-members" call to replace "show members" attribute.
(2) Component names of a PVR to be maximum of 31 wide
characters.
(3) Complex names (e.g. a partition of a partition of a PVR)
will be passed a string with the part of the names separated by
blanks ("blank" is a discussible item).
(4) Use handles for every object of the PVR except cartridges.
(5) Global names (e.g., of PVR), maximum of 4095 (wide)
characters.
(6) All tasks will be asynchronous: tasks will be created only
for mount, dismount, inventory, flush, scratch and stage. These
will:
- get back a task handle
- the task will have a wait function
- the task will have a query function
(7) Change the "pvr_area_move" back to separate functions for
"pvr_area_scratch" and "pvr_area_stage".
Action Item: Joel Williams will update and re-present his
document based on these recommendations.
The session adjourned at 9.58.
The session reconvened at 10.14
7.2 Media Partitioning
Carl Braganza initiated a discussion to deal with the following
PVL-related problem:
Cartridges (disks and/or tapes or other conceivable media)
can be divided into "sides" (for the sake of a word), and these
into "partitions" (for the sake of a different word), each of
which may represent one or more "volumes" (meaning the unit with
which the PVL deals). This structure may have further levels of
division. The resulting "volumes" may be fixed in size or
variable in the course of use.
There is a need to specify a method of describing such
structures, together with the joint and exclusivity relationships
of mount and access (i.e., whether access to one "volume"
excludes access to the other volumes, or whether multiple volumes
may be accessed simultaneously, say, hypothetically a video disk
in a special drive that has a head assembly for each side).
There is also a question how, with respect to these
structures, the user is to specify the desired pattern of use, to
that the PVL can make the right choices of assignment of volumes
to "parts" of cartridges.
Subsequent discussion showed that there are implications
here of the operations needed in the MVR, as it may have to move
between "volumes", without intervening mounts, dismounts, loads
and unloads (as currently defined). It is interesting to note
that "formatting" of media is currently (in the Reference Model)
defined to being an operation that is accomplished outside the
MVR.
Rich Wrenn noted that there appear to be at least two
requirements to satisfy:
(1) A client of the MVR wants to do the partitioning; and,
(2) A need to be able to consolidate legacy media is a way that
the an application can use the consolidated data as though it was
on the original media.
The tenor of the subsequent discussion was that there needs
to be a clear definition of which levels of the this possible
hierarchy belongs to which component of the reference model,
coming down to whether the client of a MVR should or should not
be able to see a sequence of "partitions", or not. This would
require a flag to be passed by the PVL to the MVR as to whether
it should allow the client to select "partitions" (i.e.,
"partition" is a lower level division than a "volume", although
they could be forced to be treated equivalently).
Carl asks also how this maps to request for scratch space is
another question.
Rich Wrenn listed five cases that need to be dealt with in
terms of what operations can take place, and the relationships
between the styles of partitioning.
Meeting adjourned at 11.30.
Meeting reconvened at 12.04.
Bruce Haddon opined that case (1) above (a client of the MVR
wants to do the partitioning) is only a concern of the client,
and is consequently neither a problem nor a solution.
The final consensus of the discussion was there needs to be
further information on what hardware designers are likely to do
(although others felt this was backwards, that the logical needs
of storage systems should be understood, and the hardware
designers asked to implement corresponding mechanisms, but this
was an unlikely scenario in this universe).
Action Item: Carl will attempt to collect hardware design
information, and think about making a more detailed proposal on
partitioning structures.
7.3 Time
Dave Skinner introduced a question relating to the representation
of time, and the use of time in PVR and PVL operations. This led
to a discussion of a number of scenarios proposed by Bruce Haddon
(backup windows, expensive experiments, other fixed scheduling).
The role of guaranteed times was then discussed, as well as "pre-
emptive" mounts, by Carl Braganza and Rich Wrenn.
The discussion ended with the thought that the existing
API's should be re-examined with these scenarios representing
requirements, but no specific action item was assigned.
7.4 Robotics API Proposal
Dave Skinner made a brief presentation on the role of a robotics
interface in the design and building of portable, or generic,
PVR's. This could make more libraries available with the IEEE
SSSWG model solutions. He proposed three levels of robotics
access:
- basic (information to support mount/dismount)
- intermediate (exposing more of the geometry, for the
purpose of appropriate optimization).
- full (sufficient information to, for example, create a GUI
representation of the actual physical layout of the
library and position of the robot).
Action Item: Dixon Hutchinson undertook to check whether
Exabyte would support making effort available to work on this. If
so, SSSWG will support having a PAR approved.
8 Review of Action Items
======================
(1) Contact Creig Humes to find out what there is available for
the MVR document, and make it available somehow on the web page.
Responsible: Dave Skinner.
Due: April 15, 1996.
(2) The other material from this meeting, including the API
drafts (other than MVR), and the minutes from this and past
meetings to be moved to the web page, or to be pointed to by the
web page and be available by FTP.
Responsible: Jack Cole
Due: April 15, 1996.
(3) A summary of some part of the presentations made by Rich
Wrenn at the January meeting will be prepared, and the results
made the results available on the web page
Responsible: Rich Wrenn.
Due: April 15, 1996.
(4) Investigate an appropriate site for the July meeting.
Responsible: Rich Wrenn.
Due: May meeting, 1996.
(5) Investigate the practicality of meeting concurrently with the
Goddard MSS.
Responsible: Joel William.
Due: May meeting, 1996.
(6) Investigate feasibility of combining the November meeting
with COMDEX.
Responsible: Dixon Hutchinson.
Due: May meeting, 1996.
(7) Offer a rationale for having a MGT group, and an outline for
an approach to the activities. Bob Baird will consider
contributing.
Responsible: Dave Skinner.
Due: May meeting, 1996.
(8) Propose a resolution, specifying what to advertise to whom,
and where, with respect to methods of gathering information for a
revision to the Version 5 Reference Model.
Responsible: Dave Skinner.
Due: April 15, 1996.
(9) Make text of Joel's available on the ftp server.
Responsible: Joel William.
Due: March 31, 1996.
(10) Joel Williams will update and re-present his document based
on the seven recommendations resulting from the Wednesday
discussion.
Responsible: Joel Williams.
Due: May meeting, 1996.
(11) Organize that someone make a MVR presentation, and present
an up-to-date list of issues.
Responsible: Joel Williams.
Due: May meeting, 1996.
(12) Collect hardware information about partitioning,
particularly the hardware possibilities, and think about making a
more detailed proposal on how to handle them.
Responsible: Carl Braganza.
Due: May meeting, 1996.
(13) Check whether Exabyte would support making effort available
to work on a PAR for a robotics-level interface. If so, SSSWG
will support having a PAR approved.
Responsible: Dixon Hutchinson.
Due: April 15, 1996
9 End of Session
## Resolved: that Dixon Hutchinson, and Exabyte to be thanked
for organization and support of this meeting. The arrangements
were excellent and helped considerably with the flow of the
meeting.
Moved: from the Chair.
Passed: by acclamation
Meeting adjourned at 13.41
SSSWG Home