Minutes of 3^rd EVN-NREN Meeting

Exchange Avenue, Schiphol Airport, Netherlands.

28^th January 2005 (11:00-16:30)

Present:

Name e-mail address Organisation/Institution

Walter Alef alef@mpifr-bonn.mpg.de Max-Planck-Institut

Paul Burgess pb@jb.man.ac.uk Manchester University

Simon Casey scasey@jb.man.ac.uk Jodrell Bank

John Chevers john.chevers@dante.org.uk DANTE

Michael Enrico michael.enrico@dante.org.uk DANTE

Tony Foley foley@astron.nl ASTRON

Mike Garrett director@jive.nl JIVE

Marcin Garstka marcinga@man.poznan.pl PSNC

Richard Hughes-Jones r.hughes-jones@manchester.ac.uk Manchester University

Giuseppe Maccaferri g.maccaferri@ira.cnr.it CNR-IRA

Marco Marletta marco.marletta@garr.it GARR

Ari Mujunen amujunen@cc.hut.fi Metsahovi Observatory

Michael Olberg olberg@oso.chalmers.se Onsala Space Observatory

Niels den Otter otter@surfnet.nl SURFnet

Steve Parsley parsley@jive.nl JIVE

Eugeniusz Pazderski ep@astro.uni.torun.pl Torun

Jouko Ritakari jr@kurp.hut.fi Metsahovi Observatory

Toby Rodwell toby.rodwell@dante.org.uk DANTE

Ralph Spencer res@jb.man.ac.uk Jodrell Bank

Arpad Szomoru szomoru@jive.nl JIVE

Apologies

Duncan Rogerson duncan@nosc.ja.net UKERNA

Sergei Pogrebenko pogrebenko@jive.nl JIVE

Hanno Holties holties@astron.nl ASTRON

Per Nihlen per@nordu.net NorduNET/SUNET

Agenda

11:00   Open Meeting, Introductions, Agenda and Actions

11:20   Review of overall project progress, related projects and forthcoming FP6 call for proposals (Steve Parsley, JIVE)

11:50   Detailed technical results of Jumbo Frame Tests, network monitoring and centralised control of network connection (Arpad Szomoru, JIVE)

12:20    Update from NRENs and Observatories

12:50   Lunch

13:20   New UDP Protocols for VLBI file transfer (Jouko Ritakari, Metsähovi Radio Observatory, Finland)

13:40   Bandwidth Test Results (Richard Hughes-Jones, University of Manchester)

14:00   High Bit Rate Transfers to JIVE (Simon Casey, Jodrell Bank)

14:20   Update on GÉANT2 (Michael Enrico, DANTE)

14:40   PERT and performance monitoring (Toby Rodwell, DANTE)

15:00   Discussion

16:00    AOB

16:30    Close

Note that slides from the presentations will be made available on:

http://www.evlbi.org/evlbi/evlbi.html

Minutes

The details of the presentations may be found in the slides published on the EVN website. The principal points raised in discussion were as follows:

Status of Observatories

• WA explained that connectivity to the Effelsberg observatory has proved difficult to procure. Deutsche Telecom have quoted a very high price for this link. Several options are open to achieve connectivity.
• Effelsberg have applied to be part of the VIOLA testbed program sponsored by DFN
• A grant from the Max Planck society could supply the Euro 2M required for dedicated fibre.
• T-Mobile are digging a trench in the region, in any case, which may also allow Effelsberg fibre to be laid.
• It may be possible to procure a link using cash from the IST call (See SP presentation)
• GM stated that it is hoped that Medicina Observatory will be connected by early summer. GARR are to act as consultants for the lighting of fibre in the local loop.
• The eMerlin project hopes to connect all six sites using a dedicated network by the end of the summer. It is possible at present to conduct multiple telescope observations, but only at low bandwidth. It is not possible to run Ethernet over the eMerlin network, so it would be non-trivial to connect, for example to JIVE.
• Recent problems with Onsala connection have been traced to the Mk.V processor.
• Torun observatory has a 10Gbps link to Posnan, from where there is a 10Gbps link to GÉANT. Traffic flow is limited, however as only one Mk.Va is available.

Schedule of eVLBI for 2005

The schedule was agreed (see SP presentation). A possible additional date mentioned by MG is in late June/early July for a demonstration to an EC delegation visiting Dwingeloo, including the commissioner.

The ‘Science Demo’ scheduled for 11/3 will explore dynamic scheduling- testing several astronomical sources for suitability before observing the brightest (‘flaring’) object in more detail.

A test was proposed by RH-J involving simultaneous disc-disc transfer into JIVE from several observatories. DANTE, GARR, SURFnet and PSNC confirmed that they are happy to support tests, including UDP up to 512Mbps, without prior notice. RS expressed doubts about a long (circa several hours) test over Network NW.

Action: ME to confirm that such tests are acceptable to the NRENs not represented.

EC Call for Proposals

Comments made on the presentation were:

• RH-J commented that any infrastructure should be allow for future 1Gbps++ transfers
• JR commented that lightpaths and QoS are unnecessary, in his view, as he has so far experienced no lost packets attributable to the network.
• RH-J suggested letters of support from other EC funded projects such as EGEE
• RS asked if the Square Kilometre array shouldn’t be included
• MG stated that the minimum usable bandwidth for eVLBI is 32Mbps

Jumbo Frame Tests

Comments made on the presentation were:

• RH-J commented that disc-disc transfers are poorer than expected, this points to CPU or software problems. The way in which the software is written is crucial.
• JR commented that in disc-disc transfers from Finland-JIVE the CPU usage was only 5%. The Mk.V uses considerably more.
• AS responded that US telescopes have achieved 512Mbps using Mk.V processors disc-disc.
• RH-J: Jumbo frames will reduce CPU usage but if this reduction is very significant in increasing data rates, it implies an underpowered CPU.
• WA added that new Mk.V cards will be available mid-2005.

New UDP Protocols for VLBI file transfer (and Mk.V alternatives)

• The existing Mk.V units are designed to emulate tape-based systems in their interface with the correlator. Significant work would be required to ensure that the new PCs being developed would interface correctly with the existing correlator.
• SP commented that problems have been encountered interfacing from GE cards (copper input) and the fibre supplied by SURFnet. MM (and RH-J) commented that this can be overcome using a GE converter at a cost of circa £400.

PERT and performance monitoring

• MO asked if Jumbo frames really are necessary for high bandwidth. The consensus was that this is the case if Mk.Vs are used.
• SP proposed that the next phase of tests should focus on network monitoring. Some felt that this need comprise only an element of the tests. NdO offered statistics from SURFnet.

Action: ME to check with NORDUnet etc if statistics are also available.

Action: MG to confirm if PSNC stats are available.

• MG asked if Mk.V processors can live up to the requirements of the proposal. If not, this must be decided very quickly (within 4 weeks!).

General Actions

Action: DANTE to explore monitoring possibilities

Action: SP to oversee arrangements for science demo

Next Meeting

It was proposed that the next EVN-NREN meeting be held in Dwingeloo over 2 days, in early September.