Minutes of May 9, 1996 SSC/SMC working group
tedl at exabyte.com
tedl at exabyte.com
Mon May 13 14:37:29 PDT 1996
* From the SCSI Reflector, posted by:
* tedl at Exabyte.COM
*
Document: 96-174r0 (text version)
Fort Lauderdale, FL
Accredited Standards Committee
X3, Information Technology
To: Membership of X3T10
From: Edward Lappin
Subject: Minutes of SSC/SMC Working Group Meeting
Fort Lauderdale, FL May 9, 1996
Agenda
1.Opening Remarks
2.Approval of Agenda
3.Attendance and Membership
4.SSC Topics
4.1 - Write Protect (different levels)
4.2 - Read Reverse option
4.3 - Immediate operations and reporting
4.4 - Test Unit Ready and deferred error reporting
4.5 - Capacity selection vs disk
4.6 - FCP issues
4.7 - SSA issues
4.8 - Cleaning Log Page
5.SMC Topics.
6.Other Topics
7.Meeting Schedule
8.Adjournment
Results of Meeting
1.Opening Remarks
Ted Lappin, the SSC Technical Editor, called the meeting
to order at 9:10 a.m., Thursday, May 9, 1996. He thanked
Adaptec for arranging and hosting the meeting.
As is customary, the people attending introduced
themselves and a copy of the attendance list was
circulated.
2.The draft agenda was approved.
3.Attendance and Membership
Attendance at working group meetings does not count
toward minimum attendance requirements for X3T10
membership. Working group meetings are open to any
person or organization directly and materially affected
by X3T10's scope of work.
The following people attended the meeting:
Name S Organization Email Address
Mr. Charles Monia P Digital monia at shr.dec.com
Mr. Edward Lappin P Exabyte Corp. tedl at exabyte.com
Mr. Ralph Weber P ENDL Associates roweber at acm.org
Mr. Pete Tobias A Tandem tobias_pete at tandem.com
Mr. Doug Hagerman A# Digital hagerman at starch.enet.dec.com
Mr. Gary Stephens P FSI 6363897 at mcimail.com
Mr. George Penokie P IBM gop at rchvmp3.vnet.ibm.com
Mr. Ken Hallam P Unisys ken.hallam at mv.unisys.com
Mr. Bill Dallas A# Digital dallas at zk3.dec.com
Mr. John Scheible A IBM scheible at vnet.ibm.com
Mr. Bill Martin A# Hewlett Packard wrm at core.rose.hp.com
Mr. David Cressman V Quantum dcressma at tdh.qntm.com
Mr. Roger Cummings O DPT cummings_roger at dpt.com
Mr. Dan Strevey V Intellistor/ dstrevey at intellistor.com
Fugitsu
14 People Present
Status Key: P - Principal
A,A# - Alternate
O - Observer
L - Liaison
V - Visitor
4.SSC Topics
4.1 - Write Protect (different levels)
Extensions to write protect to add new scopes and
reporting. In particular, write protect based on the
following mechanisms:
Write Protect Type Implementation ASC/Q
Hardware Exists 27/01
Software Exists 27/02 (new)
Associated (until unload) New 27/03
Persistent (record on medium) New 27/04
Permanent (not erasable) New 27/05
Software write protect needs statement on persistence (if
not already there).
Associated is kept until unload or poweroff. General
feeling that reset should also kill it (original proposal
has reset maintaining the write protect).
Persistent write protect is remembered on a media unit
basis. Questions about what happens when the cartridge
is placed in older drives.
Permanent cannot be written or appended to until
externally erased (not writable by any commands on the
target supporting the full capabilities of the media).
An ASCQ of 00h is any reason for write protect.
Action to check with Gerry Houlder if disks want this
functionality.
4.2 - Read Reverse option
Option to reverse bytes in read reverse. Problem with
original command option requiring reverse byte flow. If
make support of old form (reverse data) required but new
hardware cannot support the old form. How can this be
resolved? One way would be to have a mode bit or maybe a
new command.
4.3 - Immediate operations and reporting
Current behavior makes it difficult to determine when an
immediate operation has completed. This proposal
requests particular behavior using sense key specific
time-out values in request sense for all immediate
operations. Need to add request sense polling
information in each immediate command or in the model for
immediate commands. May want a new ASC/Q for NOT READY,
UNLOAD IN PROGRESS.
4.4 - Test Unit Ready and deferred error reporting
Should deferred errors be reported by Test Unit Ready?
Text should be placed somewhere indicating whether or not
deferred errors are allowed on TUR commands. No mode
page for selecting operation.
4.5 - Capacity selection vs disk
Do we need short tape capacity selection? Could do this
using single partition format with a percentage. Bill
Dallas thinks that we should consider using the same
method as disk drives.
Gary Stephens will generate proposal documents for items
4.1 through 4.5.
4.6 - FCP issues
Doug Hagerman presented FCP issues regarding the profile
for fibre channel SCSI devices. A private loop profile
has been started, originally for disks.
Basic problem with FCP is that no frame-level error
correction code is in fibre channel, only an error
detection code. A SCSI command maps onto an exchange.
Error checking is on an exchange with is a sequence.
This mechanism is part of the profile. Initial
justification for this mechanism came from a better error
rate using optics but the error rate has been pushed up
by speed and copper. The mechanism can create command
time-outs. The fibre profile was built on the
expectation that commands can be retried.
For performance reasons, the ACK of each frame was
removed (using class 3). Instead, the status of the SCSI
command implied reception and completion.
Ordering is maintained unless a fabric (using switches)
occurs. In the loop profile, delivery is in order for
frames. But, a missing frame is a possibility. You will
know that you are missing a frame in a sequence but not a
missing sequence.
For a data sequence, each sequence is numbered uniquely
within the command. Not clear on what the sequence
number requirements for the profile.
Question: what about the last sequence with SILI on
read?
Next is the question on how streaming and buffers work.
Doug talked about cleaning up tape operation when we
change to fibre.
If you get an error on a write command data transmission
error, the suggestion is that the command be resent. The
detection occurs during a data sequence.
However, if all of the data cannot be stored in the
drive, there is a problem. Data is written to the media
but the command has not completed. This requires that
the drive reposition to the start of the command ready
for the replacement write.
Getting into discussion about what recovery is required.
For data transmission errors, tapes could reposition to
allow a retransmission of the same command. This is not
exactly efficient. Requires that overwrites at any point
are possible.
Still, the major problem is the loss of commands. Does
not fix the problems for printers and communication
devices.
A poll indicates that typical applications use up to 64
kbyte commands.
You can always ask for the data again. This works if
something like the drive wants a retransmission (e.g.
expansion of data during compression and wants to process
the same data uncompressed) but no error occurs.
Roger would like some tape people to come to the loop
meeting to indicate the current error recovery. There is
a meeting in Santa Fe, probably Tuesday. Need to get
tape vendors to know about this meeting.
4.7 - SSA issues
John Scheible presented conversion between parallel and
SSA. Goal is to minimize conversion from parallel SCSI
while gaining performance.
SSA does have frame level retry. A disk drive target
implementation was able to reuse 85% of the code from
parallel SCSI.
Ordered delivery is guaranteed by using a single path.
Confirmed status has been added.
SSA is pipelined but not interlocked. There is no end to
end ack but there is an ack for each hop.
Contingent Allegiance is supported by locking until the
driver is notified. A temporary ACA is generated by the
protocol layer.
Commands must use the same path if order is to be
maintained.
For hot plug, the path can be rerouted by the protocol,
transparent to the command level. This is possible due
to the confirmed status.
4.8 - Cleaning Log Page
General agreement that a suggestion on the number of
bytes for the cleaning log parameter is required in SSC.
Ken Hallam wants to set the size of parameter 100h to 2
bytes (96-155r0).
5.SMC Topics.
No topics were discussed.
6.Other Topics
No other topics were discussed.
7.Meeting Schedule
The next meeting of the SSC/SMC Working Group will be in
Colorado Springs, hosted by Symbios. Actual meeting time
will be determined by the X3T10 Plenary.
8.Adjournment
The meeting was adjourned at 12:50 PM. on Thursday May 9,
1996.
More information about the T10
mailing list