JIVE Wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
==== Minutes of NEXPReS MS601 telecon on 2011-03-31 ====

The telecon was held on 2010-03-31 and started at 12:00 UTC.

** Participants: **
Simon Casey (OSO), Neal Jackson (UMAN), Jimmy Cullen (UMAN), Raplh Spencer (UMAN), Wouter Huisman (SURFnet), Hans Trompert (SURFnet), Peter Hinrich (SURFnet), Fredrik Pettai (NORDUnet), Tasso Tzioumis (ATNF), Paul Boven (JIVE).

** Agenda **
  - Introduction of newcomers
  - Status MS601
  - Status of dynamic lightpaths
  - Status of network testing task
  - Aveiro workshop
  - Evo for future telecons?
  - A.O.B.

** Minutes **

__Introductions:__

Jimmy Cullen has just started at Univeristy of Manchester, and is mostly stationed at Jodrell Bank. He previously did software programming in industry, was a grid/hepf sysadmin. In NEXPReS, his main involvement will be in WP6/task 3, testing of BoD paths.

Fredrik Pettai from NORDUnet replaces his colleague Frank Blankman. Fredrik will support the investigations into international BoD solutions.

JIVE, together with ASTRON, have interviewed several candidates for the positions that we have available in NEXPReS and hope to fill the advertised positions soon.

__Status MS601:__

Work on the document is ongoing but progressing slowly. One of the reasons for this is that the GLIF NSI protocol specification has only just been released, and that the Geant GN3/SA2 task is likewise only just getting up to speed. We've just received a list of the countries that are planning to be part of the initial GN3 BoD implementation.

__Status of Dynamic Lightpaths:__

We are in discussion with SURFnet regarding a first dynamic path between OSO and JIVE. The current plan would be to have the path be dynamic between Dwingeloo and Amsterdam, but a static VLAN path using existing cross-border connections between SURFnet and NORDUnet. One of the issues at this moment is that OpenDRAC does not (yet) have support for the Netherlight switch to make a dynamic path over longer distances.

In the Automated GOLE (GLIF Optical Lightpath Exchange) demo at GLIF last year, SURFnet and other GLIF members demonstrated automated provisioning of Bandwidth-on-Demand through a Fenius-based Superclient. A demonstration based on NSI is scheduled for the GLIF meeting in October 2011. The NEXPReS BoD applications will also be superclients that will use NSI to talk to the different BoD systems. One of the implications of this is that the NEXPReS superclient will have to perform path-finding in the BoD domain, although the initial implementation of that (with just a few available paths) will be quite trivial. Global (inter-domain) pathfinding is currently not a feature of NSI. Each NREN level BoD system will generally perform its own internal pathfinding, so that the Superclient only needs to stitch together the path through the various exchange points. NSI will eventually support heterogeneous pathfinding as an option, but will probably keep allowing the Superclient to do its own pathfinding too. It would be useful if NEXPReS were to get involved in the NSI definition and implementation process.

As the NSI protocol has been released and is being implemented already, the NEXPReS system design will skip building an implementation that uses the interim Fenius protocol, and will target the NSI protocol straight away.

__Status of Network Testing task:__

There has been a lot of activity in this task at Jodrell Bank. There is now an 80km fiber link (Jodrell Bank - Manchester) in operation that is being used for network testing. With two IBobs running the inettest firmware, a sustatined 9.95 Gb/s can be sustained in both directions simultaneously. There is occasional packet loss, which is currently being investigated. There are also 2 PCs in use for network testing, using the UDPmon software. Achieved throughput so far is 2.8Gb/s in one direction, but in the other direction the transfer speed is not stable and quite low. As the PCs used for testing aren't the same, differences in performance were expected. Testing will also encompass iperf, nttcp and netpipe in the future. Intel 10G cards as recommended by the design document from NEXPReS WP8 have also been ordered. At this stage, FPGA-based testing on the Ibob clearly gets a higher performance, but the drawback is that any changes to the current firmware will need to be done by someone well versed in VHDL. PC-based testing is much easier to set up, but at this time it is unclear whether we can indeed generate sufficient volumes of traffic to emulate the e-VLBI traffic.

__Aveiro Workshop:__

The workshop "The Growing Demands on Connectivity and Information Processing in Radio Astronomy from VLBI to the SKA" will be held from May 24-26th in Aveiro, Portugal. The program includes the EVN-NREN meeting on day two. All participants of WP6 are very much encouraged to attend this workshop. See also:
http://www.av.it.pt/workshops/ict_vlbi2ska.html

__EVO for future telecons:__

We are using the EVO system from Caltech for telecons for a number of projects, including Uniboard and NEXPReS WP8. The advantages of this system are that it supports text-mode chat as well as video chat, and is free. We will try to use EVO for our next WP6 telecon. Please check it out at: http://evo.caltech.edu/

__A.O.B.__:

None