ForCES Working Group, Monday November 4, 2013(Vancouver, Canada)Time: 1300-1430 PST Physical Coordinates: Plaza C Charter: http://datatracker.ietf.org/wg/forces/charter/ https://datatracker.ietf.org/doc/charter-ietf-forces/ Documents: http://datatracker.ietf.org/wg/forces/ meetecho: http://www.meetecho.com/ietf88/forces/ Slides: https://datatracker.ietf.org/meeting/88/materials.html#forces Jabber: forces@jabber.ietf.org Log: https://www.ietf.org/jabber/logs/forces/2013-11-04.html Audio: http://ietf88streaming.dnsalias.net/ietf/ietf883.m3uWG URL: http://tools.ietf.org/wg/forces/ CHAIRS: DJ damascene.joachimpillai@verizon.com Jamal Hadi Salim Drafts (tgz): https://datatracker.ietf.org/meeting/88/agenda/forces-drafts.tgz Drafts (pdf): https://datatracker.ietf.org/meeting/88/agenda/forces-drafts.pdf IETF88 Agenda: http://tools.ietf.org/agenda/88/#forces https://datatracker.ietf.org/meeting/88/agenda/ Materials: https://datatracker.ietf.org/meeting/88/materials.html#forces Minutes: http://tools.ietf.org/wg/forces/minutes Agenda: http://www.ietf.org/proceedings/88/agenda/agenda-88-forces https://datatracker.ietf.org/meeting/88/agenda/forces/ http://tools.ietf.org/wg/forces/agenda?item=agenda-88-forces.html DJ went over agenda and working group status: Agenda: * Administravia - Chairs (5 minutes) * WG Status -Chairs (10 minutes) * Charter Work (presentations): - draft-haleplidis-forces-model-extension (Evangelos Haleplidis - 10 minutes) - draft-haleplidis-forces-packet-parallelization (Evangelos Haleplidis - 10 minutes) - Subsidiary Management (Bhumip Khasnabish - 10 minutes) - draft-jhs-forces-protoextenstion (Jamal Hadi Salim - 10 minutes) * Non-Charter (Bhumip Khasnabish - 10 minutes) ForCES-based I2RS for E2E Control/Maintenance of MPLS (Evangelos Haleplidis - 10 minutes) ForCES-base PGW Prototype Implementation Work pre-bitsnbites demo - Remainder of time .. * Wrap up and adjourn - Chairs (5 minutes) Administrivia and WG Status (13:04): Status slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-0.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-0.pptx - Progress: RFC 6984 published and CEHA is making good headway - Lack of progress: WG is 2-3 months behind schedule - Announce ForCES demo on Thursday 19:00 at Bits and Bytes ForCES Model Extension (13:08): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-3.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-3.pdf Draft: http://tools.ietf.org/html/draft-haleplidis-forces-model-extension-04 Evangelos Haleplidis -Finalize the draft. -LFB properties issued resolved - use component id 0 for properties. -Use of bitmaps was discussed "Bit map is not a bit mask" -Jamal: Common operation to perform, so there is reasonable "need" -Very hard to do that in the current architecture -Do we need special syntax? -One approach: Setting a selector and value -Joel: It would be hard to do this because we want "mirroring" between SET/GET - and they cant be in this case. -Consensus: Model extension draft should remove bitmaps Forces Packet Parallelization (13:17): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-1.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-1.pdf Draft: http://tools.ietf.org/html/draft-haleplidis-forces-packet-parallelization-04 Evangelos Haleplidis -Last update all comments have been incorporated. -One document issue raised by Dave Hood is outstanding. -If parallel processes are available with multiple parallel paths, -then the current metadata set is not sufficient. -If nesting parallel path is allowed the merging entity does not know whether it is encountering nesting or normal metadata on the wire. -4 possible solutions Pick: Not allow nested parallel paths -Jamal suggested to only go forward with a single path -Consensus: Start simple with a single path -Evangelos believes the document is in good shape and requested to add to the charter. Jamal felt there were no obstacles to move it to a WG item but asked the WG to read it (and promised to re-read it as well as look at it from implementation). ForCES LFB Subsidiary Management (13:24): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-6.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-6.pdf Bhumip Khasnabish -From the ForCES charter update, there is a mandate to write a draft. -Not specifying how to boot VM but configuring using ForCES. -The draft is not yet written - Bhumip looking for co-authors. -Primary CE, Standby CE, Multiple FE -There is an implementation in progress (demo at bitsnbites). -Bhumip is the only one presently involved in this work. chairs feel that there has to be a reasonable draft by next meeting or we this will be killed. There were no questions. Protocol Update (13:31): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-4.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-4.pdf Draft: http://tools.ietf.org/html/draft-jhs-forces-protoextenstion-01 Jamal Hadi Salim - There was strong resistance so far against table append. Agreed to remove table append. Jamal plans to write another document in the future to cover table append vs replace vs exclusivity. -Table Range Query looks solid from implementation experience. A single success(to get/delete) is considered an overall success.. -Extended Result TLV Add an 8-bit field for result causes Rick Taylor: 2 Qs What encoding for strings? UTF-8 Prefer integer code rather than natural language string result. -Large Sparse Data Sparse data TLV has 16 bit length. This is too small since ILVs have 32-bit lengths. There is a mismatch. Q: Convert the count from bytes to words for a "new" TLV? Joel Halpern: TLV meaning mixing is a bad idea. So answer is No, since this change of meaning is surprising. Better off: Don't solve this problem, please. Rick Taylor: Can't change length based on type. Rich G?: Length is length is length Weiming Wang: Is this a question of efficiency? Designed the ILV - may need large I. ForCES-based I2RS for E2E: Control/Maintenance of MPLS (13:45): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-7.pdf (Note: PDF title incorrect) http://www.ietf.org/proceedings/88/slides/slides-88-forces-7.pdf Bhumip Khasnabish ForCES for end/mid/core point of MPLS session Where to place the agents? Data plane and control plane Show I2RS framework diagram ForCES apps to support control/manage label/label based paths Next Steps Looking for comments and interest. Will develop a draft. Jamal: We welcome these kinds of activities to show ForCES use cases. please ping Bhumip to particapace on draft. PGW Prototype Implementation (13:50): Slides: http://tools.ietf.org/agenda/88/slides/slides-88-forces-5.pdf http://www.ietf.org/proceedings/88/slides/slides-88-forces-5.pdf Evangelos Haleplidis - Shows LTE architecture on where PGW fits - Goal is to scale PGW and add Gi-LAN network functions (as ForCES LFBs) - GTP-c to control plane and GTP-u to datapath - Control app create GTP tunnels and collect stats for analytics - shows packets coming in/out of PGW and how policies are applied - demo is via 5 VMs (SGW, PGW, simulation of UE, internet and controler) Sri Gundvelli, CISCO: Question about SGW How does ForCES know what to program the PGW?... A: There needs to be LFB(s). Some were defined for this case. Kalyani Bogineni, Verizon: Is there documentation? No. Parviz Yegani, Juniper: In the demo, why is there a separate SGW/PGW? Jamal: To scale PGW. Damascene: Vertically scale or horizontally scale. Parviz: You are asking for vendors to interface with internal implemention. Industry approach is to keep existing hardware and open the top. DJ: It is beneficial to be able to open things. Parviz: IETF has to be mandated by a wireless/wireline organization. Marco Liebsh: How to scale? Multiple D-plane driven by one C-plane? Yes. Kostas Pentikousis: Horrible marketing. Don't market as this is way to scale, because the experts already know how to do this. Flexibility, yes. Demonstrate flexibility and applicability in many contexts. NFV discussions should be avoided for example. Stay the course! As the Canadians said "keep calm and carry on"! Parental Advice! Kalyani: ForCES is a protocol to split control and data plane. 3gpp and mobility management: Is it applicable? This could be a potential solution. DJ: Agrees with Kalyani. Kalyani: Wants things to be nailed to some use case and clarity provided. Evangelos: If you can model it, you can control it. Hui Deng, China Mobile: How can you handle other components like Mobility and policy management - feels this is the hard part. Sri: Theres some relevant work in the mobility WG that doesnt address the control/data issues. Can be potentially useful for his work. Kostas: So far the message seems to be that you need documentation to assess for applicability. My suggestion is to proceed to such a document. Demo followed which showed - VM startup - Initialization of P/SGW - Creation of tunnel end points - some traffic flows going through - Statistics collection Per UE and Per UE-flow Sri: Tunnel interface index comes from controller? Kalyani: Is this just demo or real deployment s/ware? Evangelos spent no more than 3 weeks so this is a demo but it could be made into a real product. Bits N Bytes demo Jamal Hadi Salim -Joint Demo between Univ of Patras, Verizon, Cumulus and Mojatatu Networks Purpose is to demonstrate the power of ForCES modeling and consistent API to define arbitrary network functions on diverse physical and virtual devices (dont care if it is under/overlay). -Went over how that demo will look like -Please come to the demo, we have gifts to give away!