Changes between Version 8 and Version 9 of GEC18Agenda/GeniSdnOffering


Ignore:
Timestamp:
11/05/13 15:10:16 (10 years ago)
Author:
mbrinn@bbn.com
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • GEC18Agenda/GeniSdnOffering

    v8 v9  
    3232The goal of this session is to discuss the Software-defined networking offering in GENI. We will review a software-based approach, the current hardware-based approach and discuss requirements for experimenter tools to allocate and manage software-controlled topologies.
    3333
     34Eric Boyd of Internet2 will present current plans and expected features for the Internet2 offering in support of the GENI SDN offering.
     35
    3436Nick Bastin of Barnstormer Softworks will be presenting a model for providing a software-switch based programmable WAN to experimenters.
    35 
    36 Eric Boyd of Internet2 will present current plans and expected features for the Internet2 offering in support of the GENI SDN offering.
    3737
    3838The remaining time of the session will consist of an open discussion, including members of the GPO infra team, of the current and anticipated state of the GENI network, with an emphasis on SDN-related capabilities for experimenters.
     
    4141
    4242None
     43
     44== Summary ==
     45
     46The session consistent primarily of two presentations, one from Eric Boyd of Internet2 and one from Nick Bastin of Brainstormer Softworks. The slides for both presentations are attached to this page.
     47
     48Eric Boyd's presentation focused on the upcoming roll-out of Advanced Layer-2 Services (AL2S) by Internet2 in late Q4 of '13 and Q1 of '14 and the impact of this roll-out on the GENI developer and experimenter community. One critical feature of the SDN approach under AL2S is the inclusion of a new service called Flowspace Firewall which is intended to work something like FlowVisor in that it is a multiplexer of experimenter OpenFlow controllers and the Firewall keeps the controllers from taking inappropriate actions on the switch (too many messages, taking up too much TCAM, extending beyond the promised flowspace restrictions). The expectation is that this service will be put out as an experimental/testbed capability and once it is mature and experimenters are proven that they can work reliably in that environment, the Firewall will go 'live' and support for live experimenter controllers on an AL2S circuit will be provided. This is expected in Q2 or Q3 of '14.
     49
     50Nick Bastin's presentation focused on a proposed approach for an SDN overlay for GENI in which experimenters can build a completely encapsulated topology and control the traffic flow decisions in that network in a manner that is transparent to other users of GENI and requires no changes to existing GENI infrastructure (H/W or rack S/W). Nick proposed an approach of creating disk images containing OVS (OpenVSwitch) and then plumbing pseudo-wires from client VM's to interfaces on those VM's to the virtual switch. By then controlling the switch (using Of or some other SDN approach) the traffic routig decisions in the whole WAN can be controlled. The approach would require some encapsulation technology (e.g. MPLS labels) to transport traffic between these switches across commodity carriers, so the experimenter would be able to see traffic at the explicit OVS control points but the traversing of the broader transport network would likely be invisible to the experimenter, much as the experimenter's traffic would be opaque to the switches managing the external transport.  We talked about the desirability of such an approach and left the next steps to be determined by later discussions within the GENI community.