wiki:GEC24Agenda/OpenFlowSupportExperimenters

Version 2 (modified by mbrinn@bbn.com, 5 years ago) (diff)

--

GEC24 OpenFlow Support for Experimenters in GENI

This is a session intended to review the current set of options available to GENI experimenters to obtain and manage an OpenFlow-controlled network topology in GENI.

Session Leaders

  • Rob Ricci, University of Utah
  • Ilya Baldin, RENCI
  • Nick Bastin, Barnstormer Softworks
  • Ryan Izard, Clemson University
  • Eric Boyd, Internet2

Agenda / Details

The session will consist of a series of relatively short presentations describing different ways that GENI experimenters can obtain an OF-controlled topology, and the steps and tools required to obtain and manage such topologies.

We will review solutions that contain elements that are:

  • Hardware topologies
  • Software topologies
  • Edge control versus transport network control

We expect the session to be interactive with Q&A during each discussion.

After the talks, we will discuss the relative costs and benefits of these different approaches to different problems. We will solicit from the audience requirements and desires for new capabilities for OF-enabled topologies that can be brought to a subsequent Developer Roundtable session at the GEC.

Summary

Marshall opened the session with a short discussion of GENI “deep programming” and SDN and OpenFlow. With the decommissioning of the mesoscale, the most well-documented ways of achieving OF-enabled topologies have gone away. The goal of the session is to discuss and review pros and cons of current and anticipated approaches towards allocating OF-controlled topologies for experimenters in GENI.

Rob Ricci, University of Utah, discussed what can be provided in InstaGENI. He strongly advocated for SW solutions (OVS, VTS) rather than HW solutions unless HW is truly called for. There is a HW solution involving NAT’ing that isn’t quite a production service.

Ilya Baldin, RENCI, discussed what can be provided in ExoGENI/ORCA. He also advocated for SW solutions.

Nick Bastin, Barnstormer Softworks, described current and planned capabilities of VTS (Virtual Topology Service) for providing custom network topologies and reviewed the current advantages and limitations of VTS.

Ryan Izard, Clemson University, described his experiences in using OVS (Open Virtual Switch) to generate OF-controlled topologies and gave a brief tutorial/description on how to turn a VM into an OF-controlled switch using OVS.

Eric Boyd, Internet2, described the Internet2 plan to separate out the production network into a non-SDN transport network (with VLAN allocation) and an SDN-controlled overlay for experimentation.

There was a lot of discussion throughout these presentations discussing, among other topics, the pros and cons of these approaches, the issues with HW and SW switches, the challenges facing switch vendors.

The session ended with some open questions for discussion: how do we expand our SDN offering beyond OpenFlow (given that there are many different frameworks, and OpenFlow is showing signs of failing). Second, we want to gather all the current best practices so that everyone can know how to use GENI resources to best set up OF-controlled topologies.

Attachments (5)