---------------------------------------------------------------------

March 14, 2013
wg-sdn call

Attending:
Dale Carder - dwcarder@wisc.edu
Michael Lambert - lambert@psc.edu
kevin mayeshiro - kmayeshiro@ucdavis.edu
Grover Browning - gcbrowni@iu.edu
Dan Schmiedt - willys@clemson.edu
Bill Owens - owens@nysernet.org
Deniz Gurkan - dgurkan@uh.edu
Chris Small - chsmall@indiana.edu
Ryan Flaherty - Ryan.Flaherty1@marist.edu
Jason Parraga - Jason.Parraga1@marist.edu
David Crowe, crowed@nero.net
Matt Zekauskas, matt@internet2.edu [through 2pm]

The agenda:

1) Agenda bash
2) Anybody from ONLAB on FlowVisor development efforts?
3) Questions this workgroup is interested in answering:
- Quick guide to starting an OpenFlow testbed
- Vendors and updates on OpenFlow implementations
- IETF and ONF and NFV standardization status
- ?

Quick guide to starting an OpenFlow testbed: 
Linksys OpenWRT + raspberryPi (FloodLight) controller (inexpensive)
Short white paper on how to build an OF testbed
Brent ? build a lab - would like to help, flavors of SDN, w/ and w/o h/w, scripts
Target audience: network engineers vs graduate students vs engineer at a company
R&E scope
Eric, Dan, Brent, Bill, Deniz, Steve (volunteering to start a draft)

Ali from ONLAB: FlowVisor (new release 2 weeks ago)
New features and new API, bug fixes, overhaul
Creation of a slice: any combination of OF parameters (in headers)
Resource slicing: number of TCAM entries, some QoS parameters, indirect CPU usage, control plane traffic limitations - always abiding with OF parameters, other info from the switches? 
Chris Small: integrate with switch a monitoring module for the flowvisor (a GENI project - which one?)
We did this as part of the IU OF campus trials but used it to show the I2, NLR and Campus networks-- Code at https://github.com/InCNTRE/Measurement-Manager - It will need some changes with the verson of FV that was just released as the API changed from XML-RPC to JSON

Bill: system requirements for FlowVisor - substantial - why?
Size of the flow space dictates the sysreq (earlier version: 100k flows, 50 slices)
Now: memory req not linear with slices, but with flow space
Size of the config consumes the memory
0.8.7 --> 1 (GENI soon)

Brent: OFTest and others, performance and feature testing, available as open source?
Ed from GNOC: specific to OESS, controller + vendor pairs
Test doc that can be shared, methodologies on the testing, look at performance and functional aspects, code being open source may be considered (?), scripts with controller + switch, failure mode variations are not easy to identify (cause - effect)
Bill: controller + switch setup? Packet-in messages?
Ed: specific LADP send back to controller, others drop, no default to send all packets to the controller
Brent - Brocade: no normal forwarding? Hybrid?
Nothing other than OF is allowed, tag/label swapping, all native OF, no hybrid
Customer facing may need to have legacy AND OF (not the complete hybrid)
Hybrid definition is also in the air since the industry is shaping up - pretty important for this group to differentiate between normal OF and hybrid
Brocade and HP are completely different in how they realize hybrid
Any consolidation of these approaches?
ONF: activity on hybrid modes was shut down by boards???!!!!
Brent: Consumers: no complete OF network is possible at this time, how do we show some architectures applicable in our medium/use cases?
Management infrastructures: vendor support is variable, more refined levels of OF/legacy designation control of switches is not present yet
OF front end to L4-7 specialized boxes? Rest of the network still IP?

Latency in operational considerations on the legacy/OF 
Normal pipeline first and then OF pipeline (order of operations in a switch)

Brent: Non-OF on the path, turn off MAC-learning, OF end points (GENI demo? at GEC16?)

Bill: I get the reasoning for having MAC learning disab

Packet-in, packet-out, inject and capture at other points

multiple OF networks stitched together through VLANs over GENI, monitor by re-directing to somewhere, hard to monitor the part where it is native VLAN (not OF) - only counters, have an IP test/monitoring point

Cisco ONE: traceroute equivalent in OF? DCN experience was similar. Carrier ethernet equipment and discrepancies with what is available on the campus.

IETF: SDN research meeting - right now.
SDNRG Agenda IETF 86

THURSDAY, March 14, 2013 Thursday Afternoon Session I (Caribbean 3)
===================================================================
CHAIR(s): David Meyer   <dmm@1-4-5.net>
          Nick Feamster <feamster@cc.gatech.edu>

AGENDA

 o Administriva                                               0 minutes

   - Mailing list:  https://www.irtf.org/mailman/listinfo/sdn
   - Web:           http://trac.tools.ietf.org/group/irtf/trac/wiki/sdnrg
   - Scribe (text/jabber)?
   
 o Agenda Bashing                                              5 minutes
   David Meyer

 o What is SDN really all about? An Architectural Perspective 20 minutes
   David Meyer <dmm@1-4-5.net>

 o SDX: A Software Defined Internet Exchange                  10 minutes
   Nick Feamster <feamster@cc.gatech.edu>

 o Introduction to ForCES                                     20 minutes
   Jamal Hadi Salim <hadi@mojatatu.com>

 o Scalable Multi-Class Traffic Management in Data Center     20 minutes
   Backbone Networks
   Edward Crabbe <edc@google.com>,
   Jennifer Rexford <jenrexford@google.com>,
   Amitabha Ghosh <amitabha.ghosh@gmail.com>

 o Network Functions Virtualisation (NfV) update              10 minutes
   http://portal.etsi.org/portal/server.pt/community/NFV/367
   Bob Briscoe <bob.briscoe@bt.com>

 o A Critical Survey of Network Functions Virtualization      10 minutes
   Daniel King <d.king@lancaster.ac.uk>

 o Towards an accelerated FPMOD based on ephemeral states     20 minutes   
   Omar Cherkaoui <cherkaoui.omar@uqam.ca>

  • No labels