SSSWG - April 1994 Minutes
IEEE SSSWG Meeting
Lake Tahoe
April 19-22 1994
SSSWG held a joint meeting with Portable Operating System
Interface (POSIX) Removable Media Study Group (P1003). The
proposed agenda, prepared by Dave Skinner, included the following
topics:
. Agenda Review
. Scope/level of abstraction (Mike Peterson)
. Review coordination model, etc.
. PAR Discussion
. MVR <--> POSIX I/O Mapping
. PVL <--> POSIX Removable Media, etc.
. Operator <--> PVR Mapping
Dave Skinner presented the Coordination Model, textual highlights
of which follow:
POSIX 1244
Application Portability Application Portability
System Component Interoperability
Distributed Systems
1. Standard I/O interface to non filesystem media (including
exclusive access)
- Non filesystem media <==> removable media
2. Mounting/dismounting of media
- Unlabeled media
- Labeled media (> 1 volume)
3. Datasets on labeled media (> 1 volume)
4. Merging dataset name space and file system space
The liaison with POSIX P1003 is Kurt Everson.
The purpose of this meeting was to provide a discussion forum for
the two groups.
Dave Skinner and several other members of SSSWG met with P1003 on
Monday April 18 and developed a model for coordination of the
activities of the two groups.
POSIX P1003 is attempting to define a system-level interface for
applications.
Mike Peterson began a discussion on what type of interface POSIX
is defining. Is it one SSSWG 1244 could use? Dave Skinner stated
that they are focused on a reduced set of functions that overlaps
some of 1244. Mike would like to be sure we and POSIX understand
the level of abstraction at which we are looking.
POSIX was not interested in accessing media in distributed
systems. They were only concerned with providing an interface to
removable media from an application. The interface would include
both a callable and command line interface (CLI). This interface
would provide a set of modules to handle mount/dismount,
positioning, open, read, write, close etc. Chuck DeBaur has
written a paper on the Application Program Interface (API) for
Mount/Dismount, etc.
Dave Skinner described the five components of the Storage System
Reference Model (VSS, PVL, PVR, SSM, Mover)
Norm Aaronson (POSIX chair) then summarized the problem(s) that
needed to be solved.
Conclusions: There is a need for standards in the areas
of removable media and sequential access
media. In particular, POSIX users and
applications need:
1 - defined behavior of open, close, read,
write operations on removable media.
2 - ability to position removable media
3 - access to and ability to influence meta-
data associated with labeled media
4 - ability to mount volumes, volume sets, and
datasets with positioning) on
labeled and unlabeled media
5 - allocation of required resources with
exclusive access
6 - access limited to datasets or volumes
7 - ability to specify all resources required
in an atomic complex.
Metadata may or may not be on the media, but the application
needs access to it. The user requirements listed above need to be
in place today. A discussion ensued as to whether they will
consider nonsequential media such as optical disks.
Applications are interested in the record structure on tape. This
information is required for data interchange. The most
significant problems are related to what happens after an error -
where is the tape positioned - how does one recover? Dave Skinner
thinks there is also no POSIX way of allocating tapes for an
application. Mike Peterson suggested examining why POSIX decided
not to tackle the file allocation problem.
The PAR developed for P1003 will specify that coordination will
occur with 1244. It will also specify the areas in which the
coordination will occur.
There is no intention to change the syntax of the read, write,
open, etc. The only change will be on behavior related to
positioning after an error. P1003 plans to standardize error
recovery and, in particular, positioning after an error.
A misunderstanding occurred about the first item in the
requirements. Mike Peterson thought they were planning to open,
read, write, etc. media. They are really talking about devices
with removable media.
Dave Skinner pointed out that a suspected overlap exists with the
PVL, the Physical Volume Repository (PVR), and the mover.
There was some discussion that I/O command handling is more of a
problem for data interchange than label formats. There was a
comment that the problems are not confined to removable media.
POSIX replied that this was true, but that the user community
problems, today, are primarily on removable tape media.Bob Coyne
pointed out that there were NO formats (no label formats, no data
formats, etc.) specified in the Reference Model.
The next portion of the meeting was a discussion by the SSSWG
subcommittee chairs of the Virtual Storage System (VSS), the
mover, and the PVL.
Bob Baird (VSS chair) described VSS as providing open, read, and
write for the client. Proper media is allocated, mounted etc. It
then passes the read/write operations to the mover. It is
essentially a block addressing model with the ability to trim
bytes at either end. (This capability is a subset of the VSS
function.) The question of the tapes being proprietary then
arose. It was agreed that the tapes are probably proprietary to
the model because the mover does not assume any format, labels,
etc.
Bob Coyne wanted to ensure that the meeting addressed the overlap
between POSIX and 1244. He also pointed out that NASA is very
interested in addressing the life cycle of media and this might
be an area that POSIX could address. The life cycle of the tape
involves more than its format. The tapes must be readable for
many years.
There was some discussion about whether "positioning" meant
absolute positioning, relative positioning, or both.
A logical volume was defined as one or more physical volumes. VSS
maintains the mapping of logical volumes to physical volumes.
Bob Baird stated that an overlap exists between VSS and the
Removable Media Study Group. If the subject that the VSS
subcommittee is addressing can be helpful, Bob will be glad to
cooperate. The ability to position to file marks is still an open
issue that should be addressed. Dave Skinner pointed out that the
POSIX group may be driving requirements for the VSS. Bob Coyne
suggested that tape marks pose a significant problem for certain
kind of drives. Bob Baird suggested that an interface might exist
that would result in tape marks for some types of devices and
another solution for other types of devices. He also indicated
that there might be different requirements for reading and
writing ANSI tapes. One might be able to read any ANSI tape, but
only write a limited subset of ANSI tapes.
Creig Humes (Mover chair) described mover's function, which is to
transfer data from a source to a sink. It positions and writes
tape marks. In addition the mover supports load and unload
functions, which are similar to open and close. It is necessary
to set up two movers to read or write. The mover can be built on
top of a driver or vice/versa. What POSIX needs for tapes is a
subset of what SSSWG needs for a driver. Some devices may make
the movera part of the device. Some vendors may choose to put
movers in control units. It was determined that a coordination
point should exist between the mover and the P1003. Rich Wrenn
pointed out that requirements should be passed to both groups and
that they should be aware that PVL-MOVER can be referenced
without VSS. Larry Pelletier stated that VSS may or may not
provide all of the functions required by P1003.
Norm Aaronson recommended that coordination should occur with
POSIX and the mover to ensure that requirements are met in both
directions. The mover will be specified as a coordination point
in the PAR.
Mike Peterson (PVL chair) described the functions of the PVL as
follows:
. Receives media associated with a device regardless of
location
. Establishes a name space for the Media and verifies that
the correct media is loaded onto the appropriate drive
. Provides resource allocation
. Maps physical volumes to cartridges
Because the media is removable, the PVL applies security on the
physical volume. The PVL is completely unaware of the data that
is written on the media. The PVL writes (via the mover) and
understands the label on the tape. The mover does not understand
the content of the label - only the position of the label. Access
to the volume label can only occur through the PVL. A tar tape is
considered to be a labeled tape with a special type of label
(i.e., none).
The PVL does not support volume sets at this time, although it
does support both parallel and serial atomic mounts. The PVL
could support volume sets in the future, which is another key
area where the PAR should specify a coordinating point. The PVL
subcommittee will respond to the requirements submitted to 1244.
Mike Peterson emphasized the dangers of implementing Mount as a
primitive. Mike says that the Mount function should not be at a
low abstraction. The formation of the study group was to
understand the nature of the problems and how best to solve them.
Chuck DeBaur pointed out the document he wrote is in response to
Mike's request at the last meeting.
P1003 will start with the list, presented by Dave Skinner, and
concentrate on number 1 (Standard I/O interface to non
file-system media, including exclusive access). If the PVL has
not solved the other problems (i.e., draft of an interface),
P1003 would then move on to the second item (mounting/dismounting
of media).
Rich Wrenn restated the sequential volume set argument. Mike
Peterson recommended that even though the PVL does not yet have a
draft document, at least four vendors exist that have PVL APIs at
which the P1003 might look. The P1003 will continue to submit
requirements to which the PVL subcommittee should respond.
Rich Wrenn suggested that some of the POSIX people interested in
the PVL attend the PVL subcommittee.
Attendees POSIX p1003/IEEE SSSWG 1244
Name Company Committee/Position
Linda Tefend E-Systems PVL
Dave Donald Metrum PVR
Betty Jo Armstead Sterling Software Secretary - PVL
Don Doerner Ampex PVR
Terje Iverson Ampex PVR
Norm Aaronson IBM P1003.W/.5 Chair
Rich Wrenn Dec PVR/PVL
Ken Stuppy IBM P1003 Removable Media G.
Chuck DeBaur Fermi Lab P1003 Removable Media G.
Kurt Everson P1244/POSIX Liaison
Hans Ridder Dec P1003 Removable Media G.
Michael T. Peterson Dec IEEE 1244.PVL Chair
Ramesh Bodaph HP IEEE 1244.MVR
Larry Pelletier IEEE 1244,VSS
Craig Humes NASA LARC IEEE 1244.MVR Chair
Steve Louis IEEE 1244.MGT Chair
Tom Jefferson Sandia IEEE 1244.MGT
Carolyn Buford IEEE 1244.MGT
Andrew Hanushevsky IEEE 1244.SOID Chair
Jodie Smith STK IEEE 1244.MGT
John L. Cole IEEE First time attendee
Dave Skinner STK IEEE 1244.PVR Chair
Bob Baird HP IEEE 1244.VSS Chair
The next meeting of SSSWG will be at PNL (Pasko, Washington)
June 7-10.
Betty Jo Armstead resigned as secretary.
Linda Tefend (E-MASS) has agreed to assume the secretarial job.
Respectfully submitted
Betty Jo Armstead - Linda Tefend
SSSWG Home