Moving Towards a Coordinated Approach for GN3
Roberto Sabatino (who is DANTE’s CTO) organised this meeting, to ensure coordination between GN3 activities is more productive.
There was an obvious interest in this meeting as instead of the normal 15 registered attendees, 21 people were present.
The discussion focused on several objectives:
- to find cross-links between all the GN3 activities and tasks;
- to organise cross-activity special interest groups;
- to find areas which are not covered by GN3 activities and tasks at the moment.
Matrix of cross-interests
To achieve the first objective, it was agreed before the meeting that the 26 x 26 matrix should be filled in by the task leaders (there are 26 tasks in the GN3 project). At the meeting people discussed how to build such a matrix. It was decided that every task leader should put forward two suggestions:
- what from his task activities/topics he thinks might produce a useful input for other tasks;
- what from other tasks activities/topics he thinks might be useful for his task.
At the meeting I suggested that for JRA1 Task 1 it would be useful to have information about Control Plane developments from JRA2 Task 2. In its turn JRA 2 Task 2 asked JRA1T1 to feed back information on the management capabilities of boxes provide from different vendors, to enable them to develop a management interface to implement the path found within boxes.
Cross-activity special interest groups
Three such groups were discussed:
- 100G;
- Virtualization;
- Control Plane.
100G: the main point of the discussion was:whether the GN3 team should aim at the demonstration or production mode of 100G (both GE and OTN) links on GEANT? Most attendees agreed that the demonstration mode would be enough. The issue of whether it should be GE and OTN wasn’t discussed. Michael Enrico, who is responsible for procurement of GENAT equipment was rather sceptical about traffic demand for all-100G links within the GEANT core; he is not a believer of the “Tsunami” theory which was presented at the plenary session, and which presents the case of us being on the doorstep of a new sharp growth in Internet traffic (new video applications is one of the reasons for such a tsunami).
Virtualization: it is about cross-activity which involves JRA1Task4 (Reza Nejabati from Essex University leads it; he is also a participant of our Carrier Ethernet project) and SA2 T1-T5. It's not of great use to us, but one of the aspects of virtualization that might be of interestb would be OpenFlow specification, which establishes an open interface between forwarding and control planes within switches or routers. Mauro Companella mentioned it at the plenary session as 'promising' and it seems to be interesting for zero control plane boxes like PBB-TE or MPLS-TP as it establishes a standard interface between such boxes and external provisioning systems (centralised control plane).
Control Plane. This cross-activity involves JRA2 Task 2 (theoretical continuation of Autobahn); SA2 Task 1 &2 – implementation of control plane protocols; JRA1 Task 1&2 as technologies investigators. Afrodite Sebasti, who leads JRA2, was very active during this discussion area. She indicated a clear interest in JRA2 to GMPLS and asked other tasks to provide her group with any information about control –plane related capabilities for hardware, as her group consists of people who have mostly a theoretical and software background and don’t talk directly with hardware vendors. She also expressed the idea of pushing vendors towards certain desirable features. e.g. implement GMPLS. She mentioned the unlucky experience they had had with Alcatel on teh GN2 project, where they waited, in vain, for suitable GMPLS features. Her idea was accepted with a great deal of scepticism, especially from people involved in procurement.
Victor Olifer - JRA1/SA1 and JRA2/SA2 combined meeting, Vienna, 10th September