8:30 AM Starting Time
Geoff Thompson chair of 802.3 calls meeting to order and gives introductory remarks.
GT: Calls for minute taker. After some discussion Paul Bottorff volunteered.
The agenda was moved for approval and voted approved by show of hands
Floyd Backes nominates Steve Haddock as chairman.
Steve Haddock is elected chairman by unanimous show of hands.
SH: suggests election of an editor.
Floyd Backes suggests delaying selection of the editor. After discussion it was decided an editor was not needed for this meeting. Selection of an editor will be made after formation of a working group.
Steve Haddock reviewed the Agenda and presented on the email and WEB addresses for the study group. The group is represented on the WEB at http://grouper.ieee.com/groups/802/3/trunk_study while the email reflector is stds-802-3-trunking@ieee.org. Information on email can be found under the agenda presentation at the above WEB address. All the presentations will be posted at the WEB site.
Everett Rigsbee took the floor to thank Steve Haddock and Extreme for the meeting rooms and arrangements.
Steve Haddock reviewed the presentation list, which included Floyd Backes, Mick Seaman, Alan Dykman, Willem Wery, Anil Rijsinghani, Norm Finn, Paul Congdon, Ariel Hendel, and Geoff Thompson.
Alan Dykman withdrew his presentation.
Steve Adams declared he would present instead of Willem Wery.
Steve Haddock introduced the first presenter Floyd Backes.
Floyd Backes presents on definitions and terminology.
The term trunk has a different meaning in 802 from the intended use in multi-link trunking. Floyd suggests using Link Aggregation rather than trunk. Other terms included in the presentation are Segment, Link Segment, End-Stations, Port, Bridge Port, Subnet, Stream, and LAN Aggregation.
Link Aggregation is a group of Link Segments, of the same medium and speed that are treated as if they are all part of the same Link Segment.
LAN Aggregation is a group of LANs or Bridged LANs, of the same or dissimilar media access method, media type or speed, that form parallel paths between a pair of connected LAN stations, and that are treated as if they form a single LAN between those two LAN Stations.
Discussion on Floyd's presentation centered on the ambiguities of the word flow. Geoff Thompson points out the word flow is used for flow control and therefore may cause confusion. Norm Finn indicates the term used should be unidirectional not bi-directional. Roth Jochen suggests the use of the work stream. NF suggests the use of thread. Others mention river.
Mick Seaman presents on Link and LAN Aggregation.
MS is focusing the scope of Multi-Link Trunking to Link Aggregation with future extensions to LAN Aggregation.
In the presentation a picture showing access to MACs associated with each 802.3 MAC and a MAC associated with the Link Aggregation is drawn.
Some discussion follows on the assignment of MAC addresses in the model. Mick points out the model may be implemented with or without use of an additional MAC address for Aggregated Link. Some discussion surrounds how to steer to the Link Aggregation MAC.
Roth Jochen states that LAN Aggregation is different from 802.1d hot standby.
Steve Adams presents on Trunking Rules/Goals.
Steve Adam states the material presented is not Intel confidential in spite of the confidential statement on the slides.
Proposes limiting work to Link Aggregation not LAN Aggregation. A Link Aggregation should be treated as single logical link or port. The packet order should be maintained. The load distribution system need not be part of the standard. The MIB should have no real time counters. Auto setup and recovery protocol should be standardized.
Discussion follows an on Aggregation of multiple speed links. Norm Finn points out that Link Aggregation were some links are of different speeds, can be simply accomplished. Anil Rijsinghani speaks in favor of aggregating multiple link speeds. Mick Seaman points out the aggregation of multiple speed links is an extension that is unnecessary for now. Paul Bottorff states aggregation of multiple speed links will complicate load balancing since the flow assigned to the low speed links will have lower burst capability than the other flows. Geoff Thompson talks to traffic distribution over multiple link speeds.
Coffee Break
Return with modification in agenda adding Geoff Thompson for Bay presenting on Scope and Architecture.
Introductions
Anil Rijsinghani Scope/Model for Trunking
Requirements
Questions on why retain frame format
Do not neglect issues related to MACs but should list MAC specific
Norm Finn on Auto/Configuration Problem Statement.
Automatic Trunking Establishment Problem Statement
NF presenting on configuration protocol for Link Aggregation
NF not talking about LAN Aggregations no matter what the presentation says
Allow manual configuration, must be separable
Floyd Backes states the automatic protocol can be built on a manual system
NF clarifies saying that the manual possible without automatic
NF says quick configuration is within a second or two
Mick Seamans says quick is in 10 msec
NF says no spanning tree topology change on link down is required for reconfiguration
Configuration protocol is interoperate with old protocols
Discussion on multi media type, Roth Jochen, ?, Norm Finn
NF does not intend to exclude multi medium types, but does not intend to design them in
NF points out half duplex is useful for talking to a sniffer or existing NICs -
NF suggests two half duplex links aggregate to a single duplex link
NF clarifies that bi-directional communication vs. half/full duplex - These are different, Norm is talking about bi-directional
Steve Haddock says we are making good time
Paul Congdon presents on Load Distribution Algorithm.
Ariel Hendel presents on Goals.
Mick Seaman states we should be careful of reflective load balance.
Geoff Thompson presents on Scope/Architecture.
Discussion
Geoff Thompson - talks against launching a project for multiple MAC protocols
Mick Seamans - Points out we need both logical MAC and physical MACs. Logical MAC addresses separated from physical preclude "plug a pray" operation
Roth Jochen talking about trade-off between separate MAC for logical and physical
Norm Finn states it is not required to have separate MACs for each product
Rich Seifert -agrees with Mick about access to both single and multiple MACs
Divander Trivity wants support of multiple MAC types
Geoff talks about filtering based on DA address
Howard ? says current standard does not steer packets through MACs
Rich Seifert will clarify
Divander Trivity wants to know if frame information is carried over logical MAC
Pat Thaler - points out the MAC service interface is not being changed
Mick Seamans -802.1 dealing with multi services in MAC
Pat Thaler -802.3 behind 802.1 in MAC service definition, the 802.3 MAC is out of date
Paul Congdon -opposed to using MAC interfaces except for control
Steve Haddock takes over the floor indicating the next agenda item is the PAR.
Floyd Backes has copies of terms presentation which will be distributed before Lunch
Everett Rigsbee - Pointers on lunch options
Break for lunch at 12:17
Reconvened at 1:40
Steve Hoddock has the PAR and Geoff's scope strawman on the projection screen
Paul Bottorff speaks in favor of limiting the scope for the initial project.
Raj ? suggests widening the scope
Steve Haddock lists of objectives on an overhead slide
Other MACs?
FDX only?
Shurock ? States that point-to-point links are what is important not duplex.
Steve Adams talks to keeping the scope wider until committee a committee is formed.
Roth Jochen states different link may be an easy extension.
Mick Seamans: argues in favor of a limited scope, pointing out that if contained in the scope we will not need to discuss the topic.
Tom Dineen speaks in favor of limiting the scope to 802.3, FDX, no mixed speed
Norm Finn: I don't want to be known as the father of 2 half duplex links aggregate to a full duplex link. Half duplex aggregation adds some complexity to standard.
Discussion on limiting to FDX operation
Geoff Thompson suggests adding same speed to Steve Haddock's objectives list on overhead projector.
Steve Haddock's options on overhead
A suggestion is made to vote on the full duplex only issue first.
Bill ?: The debate surrounding FDX or HDX in an indication they both should be allowed in the scope.
Ben ?: We tried removing half duplex from .z , but ended up supporting it anyway.
Steve Haddock: says we are not designing a MAC
Geoff Thompson and Rich Seifert indicate .x managed to limit their scope to duplex only.
David Law moves "establish an objective for Full Duplex Only", seconded by Tom Dineen
Norm Finn asks to clarify difference between Objectives and Scope. Objectives are open for debate in WG, Scope can only be changed revising the PAR to 802.3.
Objection to calling question: 4 yes, 67 no, 2 abstain
Vote on David Law's motion: yes 71, no 14, abstain 3
Passed by more than 75% to establish an objective for Full Duplex only
Paul Thaler moves to put Full Duplex only in scope of PAR, Tom Dineen second
Larry asks for no on this motion
Howard Frazier says it is a big deal to change a PAR
David Law offers a friendly amendment to change motion to point-to-point
Call question: Vote 39 Yes, 50 No, 7 Abstains, motion fails to put Full Duplex Only in PAR
David Law motioned "Put in the PAR point-to-point only", second S. Chin
Paul Congdon asks for a vote on calling the question 66 yes, 2 no
Vote on David Law's motion: 80 yes, 1 no, 2 abstain
Vote by 75%
Andrew Smith asks what the definition of point-to-point is
Geoff Thompson indicates .3 has a definition point-to-point as opposed to mixing segments
Ariel asks to entertain the issue of mixed speed in objectives
M. Uri moved to put in the PAR the works "same speed", Seconded by Vikes
Paul Bottorff says it is a waist of time to develop different speed links since the majority of the market will be single speed.
Norm Finn suggest friendly amendment to the motion changing it to an objective
Norm Finn questions the use of 75% voting rule.
Vote 82 yes, 2 no, 4 abstain to make same speed an objective
Vote by 75%
Tom Dineen makes a motion to put same speed in PAR, Paul Bottorff seconds
Vote by 75% majority is: 32 yes, 50 no, 11 abstain, motion fails
Geoff Thompson's scope is being projected.
Mick Seaman asks to change the title to Link Aggregation
Steve Haddock asks for the word trunking to be removed
Norm Finn mentions that no votes have been taken on terminology
Clarification of Link and Link Segment terminology. Link Segment is the appropriate term for point-to-point.
Discussion on Scope wording changes follows.
Rich Seifert talks about other protocols(MACs).
Geoff Thompson is editing the scope that is being projected.
SCOPE
Specify a DTE to DTE logical link which consists of n parallel instances of an 802.3 point-to-point link segment. The logical link will provide the existing 802.3 service to the MAC Client.
Define the necessary management objects and protocols to support link aggregation, including identification, addition and deletion of physical links to and from the logical link.
PURPOSE
To increase the bandwidth and availability of connections between DTEs by specifying the necessary mechanisms for parallel link segment aggregation.
Walter Thirion motions to "Accept the wording of the "scope" statement for the PAR", seconded by Dineen
Change physical links to link segments in Scope.
Bill ?: points out that "logical" is still in document and is confused since "physical" is removed.
Instead of logical link we should say link aggregation.
Steve Haddock advises word smithing is not affecting the content of scope
Discussion on the word identification indicates the word implies plug and play operation.
Tom Dineen calls the question
Rich Seifert suggests taking out the sentence on "The logical link will provide the existing 802.3 service to the MAC Client".
Discussion around the issue is the .1 definition for MAC service is different from the existing 802.3 MAC service. The 802.3 does not support bridge clients.
Friendly amendment accepted
Motion for calling the question: Vote 46 yes for calling the question.
Vote on motion for scope as written: 63 yes, 2 no, 3 abstain passes
Andrew Smith points out the scope modification changes little.
Walter Thiroin makes a motion to accept the purpose wording, Norman Finn seconds
Ariel Hendel objects to the word connections since we are working in a connectionless environment. Proposes getting rid of work link.
PURPOSE
To increase link availability and bandwidth between DTEs by specifying the necessary mechanisms for parallel link segment aggregation.
Geoff Thompson calls the question.
Vote by show of hands 79 yes, 0 no, 0 abstain
Bill ?: takes a poll of the number of people in .3ac or .1. Both groups have about 30 people.
Steve Haddock: We will continue meeting at Double Tree to approve the entire PAR
Geoff Thompson takes over for Steve Haddock
GF informs the group that approval of the PAR in March requires delivering the PAR 30 days in advance to 802.3. We must finish the PAR this week to launch a working group in March.
Steve Haddock: We must complete PAR and 5 criteria today or we will conflict with .1 and .3ac. We will continue at the Hyatt until they force us to leave, then move to Double Tree.
Geoff Thompson will fill out the project information.
Rich Seifert mentions we need to decide on the title for the PAR
Steve Haddock suggests: Aggregation of Multiple Link Segments
Discussion on Criteria 1
Norm Finn asks to add a point for improved availability for criteria 1 bullet 1.
We added text: "The ability to incrementally increase the availability of link segments".
Chin motions to accept the Broad Market Potential criteria as stated. Seconded by Macleod.
Geoff Thompson suggests friendly amendment to add "??? participants from ??? companies have indicated their support for creating an interoperable standard". The text was accepted.
Vote on motion: 48 Yes, 0 No, 1 Abstain
It is suggested to use the count taken by Howard Frazier at the tutorial.
Another suggestion is to use the vote on this motion.
Discussion on Criteria 2
M: Chin S: MacLeod Move accept criteria 2
Discussion is to use refer to .1 compatible. Geoff reads .5 PAR in this.
Concern over frame order requirements for .1d bridges for multicast
Call question: 44 Yes, 0 No, 1 Abstain
Quit at 4:30 reconvened at DoubleTree at 5:00
Discussion on 3'd criteria distinct identity
Tom Dineen moves to accept the text provided by Steve Haddock, Chin seconds
Tom Mathie has issue with the word "redundant" he wants it changed to "multiple" in second paragraph.
Vote on 3'd criteria: 31 Yes, 0 No, 1 Abstain
Discussion on 4th criteria: 802.3 Technical Feasibility
Comma after "In particular"
Geoff Thompson changed "media" to "medium"
Discussion follows which concludes that 802.3 referred to in criteria is proven technology.
Add "widely deployed" before products
Tom Dineen moves to accept technical feasibility as written, Law seconds
Vote on criteria 4: 33 Yes, 0 No, 1 Abstain
Discussion on 5th criteria: Economic Feasibility
Geoff Thompson suggests adding a statement on low installation cost and upgrades
Discussion on cost suggests the current wording is fine
Consensus is that wording is fine
Geoff Thompson suggests a cost advantage from keeping currently installed links
Motion by Tom Dineen to accept criteria 5 as written, second by Payne
Vote: 33 Yes, 0 No, 1 Abstain
Steve Haddock wants to entertain a motion "to submit PAR to the 802.3 working group chair for final editing and submission to 802 exec 30 days in advance of March meeting".
Tom Dineen moves to submit PAR to the 802.3 working group chair for final editing and submission to 802 exec 30 days in advance of March meeting. Seconded by Sanderson.
Vote is: 38 Yes, 0 No, 1 Abstain
Geoff Thompson presents on the 802.3 patent policy and process. Indicating that holders of patents must submit a written letter signed by a corporate exec or patent attorney addressed to the working group chair disclosing the patent. See patents in 802.3 bylaws on WEB. Send Geoff Thompson email if you have any questions.
Discussion on objectives
2 objectives for FDX and single speed links have already been voted and accepted.
Steve Haddock asks how many will stay another hour to form objectives.
Show of hands was about 50%
Geoff Thompson can co-ordinate a meeting at the March Plenary.
Steve Haddock says we may meet on Monday morning in March
The PAR will be presented to the exec on Tuesday.
Geoff Thompson suggests meetings Tuesday afternoon or Wednesday
Steve Haddock asks for 802.1 conflicts
Due to conflicts the meeting should not use the 802.1q Monday meeting slot
802.1 people say they will not run all week so we could work around their schedule.
Simon says objectives can vary in length, suggests making a preliminary laundry list, suggests making non-goals
Geoff Thompson concurs, but suggests use of the reflector to create a laundry list.
Will not meet Monday and will request 1 day of time during the probably on Wednesday.
Floyd Backes says 130 people attended today. We should allow for 120-150 attending in March.
Talk Howard out of his Tuesday?
Voting rules for 802.3 are voters attend 3 out of 4 days to vote in 802.3
Objectives voted by 75%
Objectives not voted
Ariel Hendel's list from presentation (Ariel Hendel)
Issues
Peter asks difference between managed or manual. The discussion indicates manual is a console interface while managed is visible on the network. The comments indicate there may be little difference between the implementations of these.
Norm Finn suggests .10 as a frame fragment example.
Goals from Ariel Hendel's goals (page 2 of presentation) added to laundry list
Geoff Thompson supports only same type media type
Norm's Laundry list
Norm draws an example with picture showing how out of order happens in link up and down.
Motion to adjourn at 6:40: No objections to adjourning.
The objective list is not voted. It will be posted for review before March.