Minutes of RAID working group 12/07/93

George Penokie GOP at rchvmp3.vnet.ibm.com
Wed Jan 12 16:02:14 PST 1994

 Minutes for RAID working group 12/07/93
 Accredited Standards Committee
 X3, Information Processing Systems
                                                     Doc. No.: X3T9.2/94-010 R0
                                                         Date: Dec 07, 1993
                                                    Ref. Doc.:
                                                     Reply to: G. Penokie

 To:         Membership of X3T9.2

 From:       George Penokie & Paul Massiglia

 Subject:    Minutes of RAID Study Group Meeting (12/07/93)


 1.    Opening Remarks
 2.    Attendance and Membership
 3.    Approval of Agenda
 4.    Report on last RAID Working Group
 5.    SCSI-3 Addressing (93-161r1)
 6.    SDA States and Types (93-191r0)
 7.    SDA Commands and Mode Pages (93-192r0)
 8.    SCSI Disk Array Model (93-140r2)
 9.    Dual-Controller Overview
 10.   Action Items
 11.   Meeting Schedule
 12.   Adjournment

                               Results of Meeting

 1.  Opening Remarks

 George Penokie the RAID Study Group Chair, called the meeting to order at 8:00
 am, Tuesday, 07 December 1993.  He thanked Vitro Corporation for hosting the

 As is customary, the people attending introduced themselves.  A copy of the
 attendance list was circulated for attendance and corrections.

 It was stated that the meeting had been authorized by X3T9.2 and would be
 conducted under the X3 rules.  Ad hoc meetings take no final actions, but
 prepare recommendations for approval by the X3T9.2 task group.  The voting
 rules for the meeting are those of the parent committee, X3T9.2.  These rules
 are:  one vote per company; and any participating company member may vote.

 The minutes of this meeting will be posted to the SCSI BBS and the SCSI
 Reflector and will be included in the next committee mailing and the next RAID
 Advisory Board mailing.

 George stated that this is the 13th meeting of the RAID study group and the
 first joint RAID Advisory Board/RAID study group meeting.  The purpose of the
 group is to deal with interface issues related to using RAIDs.  The study
 group will assess the issues and then formulate a strategy for dealing with

 2.  Attendance and Membership

 Attendance at working group meetings does not count toward minimum attendance
 requirements for X3T9.2 membership.  Working group meetings are open to any
 person or company to attend and to express their opinion on the subjects being

 The following people attended the meeting:

                      RAID Study Group Meeting Attenders

          Name          S       Organization         Electronic Mail Address
 ---------------------- - ------------------------- -------------------------
 Mr. Mark Finney        O CD Int.
 Mr. Jay Mork           O CD Int.
 Mr. Doug Hagerman      O Digital Equipment Corp.   hagerman at starch.enet.dec.
 Mr. Paul Massiglia     O Digital Equipment Corp.   massiglia at genral.enet.dec
 Mr. Howard Grill       O Formation
 Mr. David Reilly       O Formation
 Mr. Bill Hutchison     O Hewlett Packard Co.       hutch at .boi.hp.com
 Mr. George Penokie     P IBM Corp.                 gop at rchvmp3.vnet.ibm.com
 Mr. Rick Thompson      O IBM Corp.
 Mr. Carl M. Conliffe   O Jaycor
 Mr. Marlin Gwaltney    O NCR Corp
 Mr. Herb Silverman     O Peer Protocols Inc.
 Mr. Joe Molina         O Technology Forums

 13 people present

 Status Key:  P   -  Principal
              A   -  Alternate
              O   -  Observer
              L   -  Liaison
              S,V -  Visitor

 3.  Approval of Agenda

 The agenda developed at the meeting was approved.

 4.  Report on last RAID Working Group

 Last RAID Working Group was held in Colorado Springs, CO.  Several members of
 the RAID Advisory Board were present.  Minutes of the Colorado Springs meeting
 were published on the SCSI Reflector.

 5.  SCSI-3 Addressing (93-161r1) Penokie

 George presented a revised version of the SCSI-3 Addressing proposal

 It was agreed that the expanded identify message be removed from the proposal.

 Next topic was a review of some written comments on the model that had been
 forwarded by Charles Binford.  One substantive upshot was a decision to add
 some clarifying language to the description of VolSel -- =0 for
 vendor-specific defaults; =1 for use of volume sets set up by the model's
 create/modify services.

 It was pointed out that reserve is a problem.  The model does not provide for
 the situation in which a device is reserved by an initiator, followed by a
 second initiator changing the mapping mode (and therefore the meaning of the
 address that the first initiator thinks it has reserved).  The discussion did
 not result in a resolution -- George is to look into the problem further.  One
 suggestion was that reserve be changed into 'reserve target' and 'reserve
 LUN', with the former blocking all access to the reserved device, and the
 latter blocking configuration operations only.

 6.  SDA States and Types (93-191r0) Penokie

 George presented a new document defining states of a SDA and the types of
 devices needed within Disk Arrays (X3T9.2/93-191r0).

 It was generally agreed that a device type needs a model, and that models are
 lacking for fans, power supplies, LEDs, operator panels, etc.  Doug Hagerman
 agreed to draft models for these.  Later, UPS, temperature indicators, cache,
 and (maybe) pagers were added to Doug's list.

 Next, a discussion on device states ensued.  There was no clear resolution on
 the 'what states are valid for what objects' table in the model, but sentiment
 seemed to favor retaining it with perhaps some thought as to what the states
 mean and which ones are valid for which objects.  George declared that he
 would not change the document unless he got some written input from other
 attendees.  One minor editorial point:  it was agreed that the term
 'available' would generally convey the desired meaning more accurately than
 'active' as used in the document.

 7.    SDA Commands and Mode Pages (93-192r0)

 George presented a new document defining new commands and mode pages needed to
 implement SCSI disk arrays (X3T9.2/93192r0).

 The first issue was the 'get geographic location' service.  After some
 discussion it was agreed that the model needs to provide 3 services:

     -Set geographic location string & replaceability

     -Get all P-LUIs and the device type of each, and,

     -Get a specific P-LUI's type and location string.

 As an aside, it was agreed that all parameter lists should be constructed so
 they were multiples of 4 bytes in length.

 8.    SCSI Disk Array Model (93-140r2)

 George presented a revised version of the SCSI-3 Disk Array Model

 Two substantive model issues arose:

     -(DEC)  recalculation  should  provide for finer granularity
     than just an entire redundancy group. It may be desirable as
     part  of  a  host-assisted recovery procedure to recalculate
     the parity on a specific range of V-LBAs. This also  applies
     to verification.

     -(DEC)  DEC  renewed  its request for a mechanism that would
     allow   initiator-directed   throttling   of   rebuild   and
     recalculate  I/O; i.e., to allow some degree of host control
     over the amount of subsystem capacity  to  be  reserved  for
     application I/O during a rebuild or recalculation.

 Discussion around spares arrived at a tentative conclusion that a "spare in
 use" state may be required instead of the current notion that a spare is no
 longer a spare when it is in use.  Doug and George will discuss this off-line.
 The question of how to remove an association of a spare with a volume set was
 answered by stating that the spare should be deleted and recreated without the
 undesired association(s).  While awkward, this is consistent with how the
 model deals with the deletion of other objects.

 Discussion around rebuild-regeneration- recalculation clarified the
 definitions of these terms.  The group agreed that the term 'protected space
 contents' should replace 'user data' throughout these discussions, as the
 model is not generally concerned with whether genuine user data is stored in
 protected space or not.

 There was discussion about the SAM meaning of the term 'object', and the
 question of whether the model needs its own definition of that term.  George
 agreed to review the SAM document and make a determination.

 It was noted that the question of whether a given service would be mandatory
 or not will be deferred until the actual commands used to implement the
 services are defined; the model will not contain a mandatory/optional state
 for services.

 The question of what state an exchanged P-LUI takes on was left open.

 The model's provision for only permitting exchange between like device types
 was removed (relaxed?).

 In Sections and, the services need to be augmented with an
 additional parameter so that either one or all LUIs can be signified.

 It was agreed that 'rebuild P-Extent' needs to be augmented with a 'rebuild
 P-PUI' service.

 A discussion of whether section 6 of the model is actually needed was not
 resolved.  George will look at it and make a determination.

 The discussion of the model ended at about 3PM.  The RAB/RSG plan is to make
 the above mentioned minor edits to the model and submit it along with the
 addressing document to ANSI for vote at its January meeting.  George expects
 that a detailed command set specification could be submitted to ANSI for vote
 by March, 1994.

 9.  Dual-Controller Overview

 Howard Grill presented a dual-controller overview and there was a brief
 discussion.  The meeting stalled a bit on procedural issues -- what must be
 presented to ANSI and how?  While the energy level was low (late in the day),
 the group agreed that the right approach would be to firm up the proposal in
 the RAB/RSG and present a more complete concrete document to ANSI.

 10.  Action Items

 a) Penokie - Revise the SCSI-3 Addressing (93-161) document and present this
 proposal to the next plenary for a vote.
 b) Penokie - Revise the SDA States and Types (93-191) document.
 c) Penokie - Revise the SDA Commands and Mode Pages (93-192) document.
 d) Penokie - Revise the SCSI Disk Array Model (93-140) document and present
 this proposal to the next plenary for a vote.

 11.  Meeting Schedule

 The next meeting of the RAID Study Group is planned for Jan 11, 1994 at the
 Catamaran Hotel in San Diego, CA.  The meeting is expected to run from
 9:00am-5:00pm.  This meeting will be a joint RAID Advisory Board Host
 Interface Group and X3T9.2 RAID Study Group meeting.

 12.  Adjournment

 The meeting was adjourned at 03:30 pm. on Tuesday, December 07 1993.

More information about the T10 mailing list