forces - Wednesday, 7 November 2012

forces

forces@jabber.ietf.org

Wednesday, 7 November 2012< ^ >

Evangelos Haleplidis has set the subject to: IETF 85 ForCES WG

Room Configuration

GMT+0
[18:51:56] Evangelos Haleplidis joins the room
[19:46:41] adrianfarrel joins the room
[19:48:01] equinox joins the room
[19:49:34] <equinox> WG status
[19:49:36] <equinox> updates since last meeting, LFBLibdraft
[19:50:08] <Evangelos Haleplidis> Has the meeting started?
[19:50:13] <equinox> yes, meeting has started
[19:50:46] <equinox> JHS> lots of renewed interest, openflow, etc.
[19:50:56] <Evangelos Haleplidis> i can't hear anything through the audio...
[19:51:03] <Evangelos Haleplidis> it's room 209 right?
[19:51:05] <equinox> yes
[19:51:11] <Evangelos Haleplidis> ouch.
[19:51:14] <equinox> no idea how to fix that, sorry
[19:51:16] <Evangelos Haleplidis> i heard a door opening.
[19:51:19] <equinox> i'm last-minute scribe
[19:51:26] <Evangelos Haleplidis> thanks for that as well.
[19:51:39] <equinox> JHS is Jamal Hadi Salim, WG chair, currently speaking
[19:52:04] <equinox> JHS> requested for more participation, lots of private mails, but not much public
[19:52:36] <equinox> JHS> 2 new drafts not in scope of charter
[19:53:15] <equinox> JHS> work as chartered is mostly complete
[19:53:36] <equinox> JHS> WG boring because unscoped work not allowed
[19:55:24] <equinox> JHS> this meeting is about discussing terms outside charter, then either extend charter or shutdown WG
[19:56:12] <equinox> slide: Forces Architecture
[19:56:19] <equinox> JHS> requests for FE<>FE plane
[19:56:27] <equinox> slide: new Work
[19:57:18] <equinox> slide: New Work [that] may not require recharter
[19:57:54] <equinox> JHS> (going through the slide items)
[19:58:12] <equinox> JHS> SCTP simplification, 3-socket communication model
[19:59:16] <equinox> JHS> IRS northbound definition, supporting multiple CEs?
[19:59:30] <equinox> JH> Joel Halpern
[19:59:36] <equinox> JH> multiple controllers controlling the same agent
[19:59:45] <equinox> JH> not high availability draft
[20:00:33] <equinox> JHS> ok
[20:01:09] <equinox> (blue sheets)
[20:01:29] <equinox> moving to first draft
[20:01:35] <equinox> ForCES Model Extension
[20:01:47] <equinox> presented by JHS, who's representiong you apparently Evangelos ;)
[20:01:59] <Evangelos Haleplidis> yes! :)
[20:03:16] <equinox> JHS> complex metadata
[20:03:20] <equinox> JHS> iteration and executing over data
[20:03:26] <equinox> (do i really need to copy everything?)
[20:03:43] <Evangelos Haleplidis> will be useful for minutes, but it's up to you! :)
[20:03:49] <equinox> JHS> different datatypes, default values
[20:03:53] <equinox> yeah i'm copying it to minutes
[20:04:29] <equinox> JHS> problems encountered when defining the language
[20:04:32] <equinox> JHS> requires model change
[20:04:50] <equinox> JHS> model was intentionally simple, but now problematic
[20:05:22] <equinox> JHS> using data as metadata and letting the user taking care of what it means (?)
[20:05:32] <equinox> Evangelos: do you want to comment?
[20:05:39] <Evangelos Haleplidis> i didn't hear the question very well
[20:05:41] <equinox> (they're asking for you)
[20:05:43] <Evangelos Haleplidis> can he speak to the mic?
[20:05:55] <equinox> there was no question :)
[20:06:08] <equinox> they just asked if you want to comment anything
[20:06:18] <equinox> next up is JH (Joel Halpern)
[20:07:00] <equinox> >> ForCES Parallelization Draft
[20:07:05] <equinox> JH> forces can not really support multiple actions in parallel
[20:07:27] <equinox> JH> need to represent what's actually possible
[20:07:48] <equinox> slide: Introduced LFBs
[20:08:11] <equinox> JH> new LFBs, splitter, merger
[20:08:46] <equinox> JH> needs more work
[20:09:04] <equinox> JH> not within charter
[20:09:14] <equinox> JH> want to fix charter
[20:09:28] <equinox> JH> not going through slides, cutting short here, just want to point out charter change
[20:10:11] <equinox> JH> list of supported LFBs tells CE what works
[20:10:13] <equinox> BK: Bhumib Kashnabish
[20:10:35] <equinox> grah, i didn't really get the question, it was about chartering
[20:10:52] <equinox> JH> yes, we're just looking what's needed on the charter
[20:11:29] <equinox> next presentation, InterFE LFB
[20:11:56] <equinox> presenter: <?>
[20:12:14] <Evangelos Haleplidis> presenter: Damascene Joachimpillai
[20:12:15] <equinox> ?> modifying packet format, adding new fields to support InterFE LFB
[20:12:18] <equinox> thanks :)
[20:12:33] <equinox> DJ> application-specific metadata
[20:13:29] <equinox> JH> bad problem, needs to be solved
JH> packet format simpler if stuff was metadata attached to packet
redirect TLD carries all metadata, would be available, information for decision
simpler format possible?
[20:13:52] <equinox> JHS> mostly based on CE-redirect packet
JHS> redefine in the same way
[20:14:13] <equinox> aaand next presentation...
[20:14:24] <equinox> if you have a question btw just drop it in i'll try to put it through
[20:14:59] <Evangelos Haleplidis> ok thanks :)
[20:15:14] <equinox> JHS giving intro before next draft
JHS> out of charter scope again
JHS> missing interfaces CEM/FEM to anything else
[20:15:34] <equinox> JHS> tell CE where FE is, etc.
[20:16:00] <equinox> JHS> implementation experience shown it useful
[20:16:50] <equinox> JHS> use case: Single NE, booting up multiple virtual FEs, want CEM -> FEM to get instances booted up
[20:17:19] <equinox> (draft is XEM interface)
[20:17:39] <equinox> JHS> needs CEID/FEID reserved
[20:18:15] <equinox> aaaand next draft (this is really express mode here...)
[20:18:44] <equinox> JHS> again pointer to discussion about recharter/shutdown
[20:19:04] <equinox> next draft: Forces on Openflow switch
[20:19:18] <equinox> presenter: Amir Montreal (?)
[20:19:25] <equinox> JHS> did that at university
[20:19:26] <equinox> err
[20:19:32] <equinox> AM> did that at university
[20:19:56] <Evangelos Haleplidis> Omar Cherkaoui
[20:19:58] <equinox> AM> evaluate flexibility between ForCES and OF
[20:20:04] <equinox> thanks again :)
[20:20:23] <equinox> Q: OF version?
[20:20:28] <equinox> OC> 1.1 and 1.2
[20:20:57] <equinox> Q was from BK: Bhumib Kashnabish
[20:21:31] <equinox> OC> looking at flexibility & behaviour of 100G OF switches
[20:21:58] <equinox> OC> not performance focus
[20:22:17] <equinox> OC> working on tools
[20:22:44] <equinox> OC> explaining mapping of OF table to set of LFBs
[20:23:18] <equinox> OC> want to be generic in which forwarding model is evaluated
[20:23:26] <equinox> OC> TCAM, SRAM, etc.
[20:23:39] <equinox> OC> also want to evaluate different applications
[20:23:56] <equinox> JSH> time constraints, slides are online, go to demo
[20:24:27] <equinox> OC> describing platforms, currently working on other board (NFP PCIe card?)
[20:25:35] <equinox> JSH> interrupting, meeting on Thursday, Salon E
[20:25:42] <equinox> JSH> setup will be running
[20:25:54] <equinox> JSH> or stay here after this session
[20:26:31] <equinox> next part - work that does not require recharter
[20:27:04] <equinox> JHS == JSH == typo :)
[20:27:11] <equinox> JSH> anyone interested in this?
[20:27:45] <equinox> BK> is IETF going to blindly endorse OpenFlow?
[20:28:04] <equinox> Adrian Farrell, Juniper> No.
[20:28:16] <equinox> BK> so we need to work on this
[20:28:49] <equinox> BK> need to look what can be delivered & expanded
BK> are implementors limited to L2? L3 concept?
[20:29:41] <equinox> ? (Google)> interesting work
academic focus but no implementations (commercial)
JSH> commercial is totally different focus
[20:30:04] <equinox> ? = Edward Crabble
[20:30:32] <equinox> JHS> problem is not that forces does not work, does pay people's lunch, there is interest
[20:30:45] <equinox> DJ (Verizon)>
[20:30:51] <equinox> DJ> looked at many shipped stuff
[20:31:07] <equinox> DJ> looking for mechanism that will actually work well in controllered datapath as well as datapath to datapath
[20:31:16] <equinox> DJ> went looking, only easy to use was ForCES
[20:31:24] <equinox> DJ> OpenFlow does not have the constructs
[20:31:37] <equinox> DJ = DJ Joachimpilliai
[20:31:51] <equinox> DJ> defining stuff in ForCES that was not readily usable in OF
[20:32:04] <equinox> DJ> will worst-case use OF as black box and put ForCES on top
[20:32:16] <equinox> DF = Dan Fross, Cisco>
[20:32:24] <equinox> DF> big vendors perspective
[20:32:32] <equinox> DF> BV should not have an impact on IETF work
[20:32:50] <equinox> DF> stuff doesn't get accepted, but maybe later in future people come back and say "we need this"
[20:32:56] <equinox> DF> lots of work put into architecture
[20:33:00] <equinox> DF> good ways to extend it
[20:33:11] <equinox> DF> lack of public eye / vendors should not limit activity
[20:33:21] <equinox> DF> future work depends on how architecture develops
[20:33:37] <equinox> EC> stuff is relatively mature
[20:33:46] <equinox> DF> people have different ideas about whether additional work needs to be done
[20:33:55] <equinox> EC> agree
[20:34:11] <equinox> EC> parallelism stuff is needed very muc
[20:34:26] <equinox> JHS> other solutions are not in list, not enough time
[20:34:31] <equinox> AF> request to name other suggestions
[20:34:40] <equinox> JHS> CE to CE
[20:34:50] <equinox> JHS> maybe impl without any changes to ForCES architecture
[20:35:17] <equinox> AF> even without WG, can write drafts & do RFCs
[20:35:33] <equinox> AF> if not enough critical mass, still fine
[20:35:47] <equinox> AF> is there a group of people wanting to work together to work on "forces next-gen?"
[20:35:54] <equinox> JHS> hard question to answer here
[20:36:05] <equinox> JHS> what would be a critical mass?
[20:36:19] <equinox> AF> how many people are tourists? (show of hands)
[20:36:32] <equinox> hands -> a few 15%
[20:36:38] <equinox> AF> starting with small base
[20:36:57] <equinox> AF> maybe not enough
[20:37:14] <equinox> AF> people who want to build and deploy?
[20:37:21] <equinox> JHS> have people who want to build & deploy
[20:37:30] <equinox> JHS> DJ wants to deploy, doesn't work, out of charter
[20:37:40] <equinox> JHS> JH's requests also out of charter
[20:37:56] <equinox> AF> don't care about charter; if everything was allowed, would it still be a WG?
[20:38:21] <equinox> JHS> get back to "is there a critical mass?"
[20:39:00] <equinox> time almost over
[20:39:07] <equinox> discussion has ended
[20:39:12] <equinox> ok closing :)
[20:41:23] <equinox> minutes will be the same as this log
[20:46:20] equinox leaves the room
[20:50:41] Evangelos Haleplidis leaves the room
[20:52:25] adrianfarrel leaves the room
[20:53:31] adrianfarrel joins the room
[21:03:26] adrianfarrel leaves the room
[22:10:52] equinox joins the room
[22:57:32] equinox leaves the room