Draft minutes of Data Integrity SG -- August 19-20, 2003

John Lohmeyer lohmeyer at t10.org
Wed Aug 20 12:09:28 PDT 2003

* From the T10 Reflector (t10 at t10.org), posted by:
* John Lohmeyer <lohmeyer at t10.org>

 Accredited Standards Committee*
InterNational Committee for Information Technology Standards (INCITS)

                                                    Doc. No.: T10/03-281r0
                                                        Date: August 20, 2003
                                                    Reply to: John Lohmeyer
To:       T10 Membership
From:     Ralph Weber and John Lohmeyer
Subject:  Data Integrity Study Group -- August 19-20, 2003
          Denver, CO

1.     Opening Remarks
2.     Approval of Agenda
3.     Attendance and Membership
4.     Old Business
4.1      End-to-End Data Protection Proposal (03-111r0) [Houlder]
4.2      CAP Data Integrity Usage Models (03-222r0) [Sheffield, Holt, Rassbach]
4.3      Simplified End-to-End Data Protection (03-176r3) [Penokie]
4.4      End-to-End Error Cases (03-207r3) [Penokie]
4.5      End-to-End Data Protection justification (03-224r0) [Holt]
4.6      Checksum vs. CRC guard algorithm
4.7      16-bit vs. 32-bit CRC
4.8      Long CDB vs. short CDB
5.     New Business
6.     Review of Recommendations
7.     Meeting Schedule
8.     Adjournment

Results of Meeting

1.     Opening Remarks

John Lohmeyer called the meeting to order at 1:00 p.m. Tuesday, August 19,
2003. He thanked LSI Logic for hosting the meeting. As usual, the people
present introduced themselves.

2.     Approval of Agenda

The draft agenda was approved with no changes.

No items was added or revised during the course of the meeting.

3.     Attendance and Membership

Attendance at study group meetings does not count toward minimum attendance
requirements for T10 membership. Study group meetings are open to any person
or organization directly and materially affected by T10's scope of work. The
following people attended the meeting:

         Name          S        Organization         Electronic Mail Address
---------------------- -- ------------------------- -------------------------
Mr. Walter Rassbach    V                            rassbach at nilenet.com
Mr. Tim Symons         A  Adaptec, Inc.             Timothy_Symons at adaptec.
Ms. Pat Thaler         P  Agilent Technologies      pat_thaler at agilent.com
Mr. Robert H. Nixon    P  Emulex                    bob.nixon at emulex.com
Mr. Ralph O. Weber     P  ENDL Texas                roweber at acm.org
Mr. George O. Penokie  P  IBM / Tivoli Systems      gop at us.ibm.com
Mr. Robert Sheffield   A  Intel Corp.               robert.l.sheffield at intel.
Mr. John Lohmeyer      P  LSI Logic Corp.           lohmeyer at t10.org
Mr. Keith Holt         V  LSI Logic Corp.           keith.holt at lsil.com
Mr. David Weber        V  LSI Logic, Inc.           david.weber at lsil.com
Mr. Mark Evans         P  Maxtor Corp.              mark_evans at maxtor.com
Mr. Edward A. Gardner  AV Ophidian Designs          eag at ophidian.com
Mr. Craig W. Carlson   A# QLogic Corp.              craig.carlson at qlogic.com
Mr. Michael Walker     V  Seagate                   michael.d.walker at seagate.
Mr. Jim Coomes         A  Seagate Technology        jim.coomes at seagate.com
Mr. Gary D. Moorhead   V  Seagate Technology        gary.moorhead at seagate.com
Mr. Steven Sletten     V  Sun Microsystems          steven.sletten at sun.com
Mr. Charles Binford    A  Sun Microsystems, Inc.    Charles.Binford at sun.com
Mr. Richard Lary       V  TuteLary, LLC             ANSI-T10 at lary.com

19 People Present

Status Key:  P    -  Principal
             A,A# -  Alternate
             AV   -  Advisory Member
             L    -  Liaison
             V    -  Visitor

4.     Old Business

4.1    End-to-End Data Protection Proposal (03-111r0) [Houlder]

This topic was discussed under items 4.2 and 4.7.

4.2    CAP Data Integrity Usage Models (03-222r0) [Sheffield, Holt, Rassbach]

The group discussed the structure and content of the Data Integrity Field
(DIF) subfields and reviewed the uses for each of the subfields based on the
contents of 03-222r0 and 03-111r0.

Jim Coomes requested a straw poll on whether the proposal should include
mechanisms for validating the contents of the Meta Tag DIF field at the
target. The straw poll favored not validating the Meta Tag at the target with
a vote count of 13:2:1.

Mark Evans requested a straw poll on whether the proposal should include a new
16 byte Read/Write CDB format with a Ref Tag field. The straw poll was against
adding the new 16 byte CDB format with a tally of 12:3:4.

4.3    Simplified End-to-End Data Protection (03-176r3) [Penokie]

George Penokie led a line-by-line review of his end-to-end data protection
proposal (03-176r3). The group discussed the structure of the Data Integrity
Field, formatting disks, legacy read commands, legacy verify commands, and
numerous other aspects of the proposal.

Ed Gardner requested a straw poll to advise George on whether discussion of
the Extended Copy command should be retained in the proposal. The straw poll
favored removing the Extended Copy 8:0.

The group requested numerous changes in the proposal and George agreed to
prepare a new revision for consideration at the September CAP meeting.

4.4    End-to-End Error Cases (03-207r3) [Penokie]

George Penokie asked that this topic be deferred to the September CAP meeting.

4.5    End-to-End Data Protection justification (03-224r0) [Holt]

Keith Holt asked that this topic be deferred to the September CAP meeting.

4.6    Checksum vs. CRC guard algorithm (03-247r0)

Walter Rassbach presented 03-274r0 showing how use of a checksum guard can aid
in detecting RAID data reconstruction (a.k.a., write hole) errors. Ed Gardner
noted that application of a vendor specific guard algorithm on only the RAID
parity drive will provide the same level of checking. Richie Lary reviewed the
error rates that the various algorithms (including checksum) will detect and
concluded that the proposed checksum usage represents a line of a defense for
an error case that should have been caught very much earlier in the activities
of a device.

Ed Gardner and Pat Thaler stated unequivocally that CRC is superior to
checksum for detecting bit flips. Walter stated that checksum should be
defined as the only guard algorithm. Richie noted that in the context of this
system, CRC and checksum have nearly equal error detection effectiveness.

Jim Coomes requested a straw poll to chose between CRC, checksum, and both for
the guard algorithm. The poll results were as follows:
o      15 - CRC
o      2 - checksum
o      0 - both

Walter Rassbach requested that the CRC have no seed. Richie Lary asked that
there be no 1's complement at the end. The group agreed with both ideas.

4.7    16-bit vs. 32-bit CRC

The group discussed the 16-bit CRC proposal presented in 03-111r0. George
Penokie requested that the CRC description be revised to conform to the CRC
description form found in 03-176r3 and/or SAS.

During the second day, Ed Gardner asked the group to consider the 1100B CRC
instead of the 18BB7 CRC.

The group agreed to study the various options on the table and revisit this
issue during the September CAP working group meeting.

4.8    Long CDB vs. short CDB

This topic was discussed under item 4.6.

5.     New Business

No new business was brought before the group.

6.     Review of Recommendations

No documents were recommended for approval during this meeting.

7.     Meeting Schedule

The next Data Integrity discussions will be held as an agenda item in the CAP
Working Group, Tuesday, September 9, 2003 commencing at 1:30 p.m. and
continuing Wednesday, September 10, 2003 from 9 a.m. to 7 p.m. (or until all
agenda items are completed). The meeting will be in Seattle, WA at the
Renaissance Hotel (800-278-4159), hosted by Microsoft.

Additional teleconference calls may be announced on the T10 reflector as

8.     Adjournment

The meeting was adjourned at 12:15 p.m. on Wednesday August 20, 2003.

* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org

More information about the T10 mailing list