SDN Meeting Notes - Artur

People present: Inder Monga, Joe Metzger, Chin Guok, Dale Finkelson, Michael Ernst, Edoardo Martelli, Lars Fischer, Magnus Bergroth, Michael Bredel, Ronald van der Pol, Artur Barczyk

Via VC: Mian Usman
(did I forget anyone?)
The discussion centered on the following notion:
The strength of SDN/Openflow lies in being a possible platform
for resolving issues for which no commercial solution
currently exists. Do we have such problems in LHCONE currently?
a) Joe: transatlantic links is an unsolved problem.
   As presented in previous LHCONE meetings, the multiplicity
   of transatlantic circuits poses a challenge in how to
   most effectively balance the flows across them.
   Also taking into account these are not dedicated circuits.
   There is no "standard way" currently to address this problem.
   Openflow could be a solution here.
   Caltech is working on a solution, through DOE ASCR funding. The
   OLiMPS project will be demonstrated at TNC, and tested first in
   US LHCNet.
   It can be tested using a mixed infrastructure, using USLHCnet,
   ACE and other circuits, with openflow switches placed at exchange
   points. [E.g. the ACE circuit links for a proof-of-principle test?]
   Current implementation relies on a single controller, but the project
   is open to collaboration on a federated approach.
b) Chin asked about use case for "reliable multipoint". There might be
   potential interest in ATLAS. Chin will continue discussion with
c) Michael Ernst inquired about Openflow for building a scalable
   virtualised data center. ATLAS is looking at an elastic cloud
   solution based on OpenStack. Virtualized data center bridging
   is certainly a good application space for Openflow, with a
   direct benefit to the experiments/sites.
   BNL has an OpenFlow based installation, and a second site will
   soon come up in the US.
   They are interested in testing this on the ESnet testbed.
   Extend to other Tier1/Tier2 sites? Interest from other sites with
   OpenStack installation?
Artur will prepare an SDN session at the next LHCONE/LHCOPN workshop
in Paris. If you'd like to present you ideas in this space, please
send me a short note.

Submitted by David Foster on Mon, 05/13/2013 - 08:02