Minutes LHCONE Point2Point Service call - February 17, 2015
Attendees: Kostas Stamos, Shawn McKee, Azher Mughal, John MacAuley,
Tangui Coulouarn, Vlad Lapadatescu, Gerben van Malenstein
Apologies: Bill Johnston
1. Going forward from Cambridge meeting
After a constructive and open group discussion in Cambridge it has been
decided in today's call to create 2 technical subgroups that will be
implementing different solutions within the upcoming two months. We
agreed that the deadline for these groups is the GLIF meeting on March
26 and 27.
Tangui: post-meeting in Cambridge: network connectivity 'seems to be
solved now', sites from ESnet can be seen from GÉANT. We still have an
open issue in the last-mile: an integration issue within the LHC site.
Shawn: KIS(S). There’s two components: 1. setting up the circuit 2.
routing site traffic into circuit, could be BGP.
Vlad: expecting to have something running (setting up layer 2 circuits
between sites) by the end of February.
- Solution Subgroup 1 (OpenFlow):
Vlad suggested using OpenFlow devices in the path at the LHC sites.
Azher: Caltech is to implement the “OpenFlow” use case (PhEDEx, CMS).
This subgroup will also take into account LHC sites that are not having
OpenFlow capabilities. Trying to work with FERMILAB (Phil DeMar),
following this up end of February. Following up now with Brazilian Tier2
partners HEPGRID & SPRACE with Brocade/Dell OpenFlow switches. Azher
willing to lead this effort. SDN workshop by Internet2 is planned end of
March in Florida ; Azher/Caltech willing to demonstrate.
- Solution Subgroup 2 (BGP):
Gerben: Brookhaven (Michael Ernst) has committed by email to join
dynamic service effort together with SURFsara. Since no people from
these sites are on the call now, Gerben will ask Sander Boele to follow
up with Michael Ernst. This subgroup will build a BGP-based solution via
dynamically created VLANs and will be lead by Sander Boele. Chin Guok
and Magnus Bergroth offered their help. From the AutoGOLE John MacAuley
and Hans Trompert are assisting with the dynamic WAN part.
Tangui: do we check if the OpenFlow and the BGP-based solution are
compatible? It's proposed to first creating the two solutions and then
compare two working solutions.
The question was raised if LHC sites strongly prefer one or the other
solution: OpenFlow or BGP. It was suggested to put this question on the
mailing list with the minutes: so, if you're part of one of the LHC
sites, could you please let Gerben know if there's any preference for
one or the other solution?
- No other business to discuss.
Our next call will be on March 10, 2015 (16:00-17:00 CET).