Minutes of NEXPReS WP6 telecon on 2012-01-11

Particpants

Fredrik Pettai (NORDUnet), Richard Hughes-Jones (DANTE), Ralph Spencer (UMAN), Neal Jackson (UMAN), Jimmy Cullen (UMAN), Simon Casey (OSO), Wouter Huisman (SURFnet), Hanno Holties (ASTRON), Yan Grange (ASTRON), Fedde Bloemhof (JIVE), Roelof Anne Schoenmaker (JIVE), Paul Boven (JIVE)

Agenda

  1. Updates from all parties
  2. Status NSI work and network infrastructure
  3. Deliverable D6.07, Demo international BoD 4Gb/s
  4. Deliverable D6.01, Proof-of-concept BoD client
  5. A.O.B.

Minutes

Opening (Paul Boven): Thank you all for joining in such large numbers, this is the largest WP6 telecon to date.

Updates from all parties

Fredrik Pettai (NORDUnet): NORDUnet held a new internal meeting, decided they will not continue development on OSCARs so they will only be able to support IDCP (AutoBahn) protocol. Instead they are now focusing on OpenNSA, which they also used for the SuperComputing testbed to provide NSI. The NSI testbed is available to be used for the NEXPReS deliverables. The NORDUnet OpenNSA can only be used for crossconnect at the moment. There is no date for moving OpenNSA into production yet, but the testbed does work for a proof-of-concept. Onsala could easily be connected behind the testbed. In Metsahovi, there was a problem with delivery of a router that has to be installed at FUNET, this is delayed until mid februari. This router is neccesary for both NSI and AutoBahn based tests.

Richard Hughes-Jones (DANTE): Update on BoD within GEaNT, they will be able to supply IDCP based BoD accross London and NORDUnet so a connection between Jodrell Bank, London, NORDUnet/FUNET and then Metsahovi, where the segments between Londan and FUNET would be dynamic. There was a large demonstration of NSI at the Future Internet Association in Poznan, and at the Supercomputing conference, both went very well. AutoBahn can now use NSI and has been shown to interwork succesfully with other NSI implementations. The NSI 1.0 document has been written, containing fixes for issues found in Rio and onwards. This will be published as an Internet Informational Document by the OGF, not yet formalized as a standard. It is mostly an NNI interface, and important features like security, topology, monitoring and directory are still missing. There is also discussion about providing an UNI (User-to-Network Interface). Version 2.0 is planned for mid-2012, and will hopefully have all the functionality for a full BoD-service. There will be further demonstrations at TNC, GLIF meetings and the next Supercomputing.

Jimmy Cullen (UMAN): They built two PC systems as per NEXPReS deliverable 8.2 and tested them, and found they can send and receive data at 9.98 Gb/s, even under load. They built a web-based fronted to use UDPmon for testing links. A 4Gb/s link from Jodrell Bank to the GEaNT Pop in Londen has been requested from Janet, this request is still in progress. Once this link is available they hope to be testing at 4Gb/s from JBO to Metsahovi. They are also performing disk and Raid-performance tests on the new test machines, as the software to be delivered in WP8 is not yet available. The system currently supports 2x 1Gb/s streaming, and 4Gb/s streaming works but has some packet loss. Next they want to try if using IO_DIRECT will improve performance. JBO have submitted an abstract for the TNC-2012 to present a paper on comparing FPGA versus PC-based network testing.

Simon Casey (OSO): Have built a similar test-PC with 24 disks and a Myricom 10G card, are awaiting 10G connectivity to the server room it is installed in.

Wouter Huisman (SURFnet): Just before Christmas, JIVE did the first succesful NSI request on the SURFnet OpenDRAC NSI infrastructure. Their testbed is connected to NETHERlight and from there connectivity towards OSO should be possible.

Hanno Holties (ASTRON): ASTRON have just hired Yan Grange to work full-time on NEXPReS. They are now getting up to speed, as most of their deliverables are late in the project. They have already done many data transfers for Lofar, but now need to start integrating BoD into that process.

JIVE (Paul Boven): The NEXPReS NSI client that is being developed at JIVE can now create reservations when talking to the SURFnet BoD-system. We can connect our port at JIVE to a testbed in Amsterdam, and from there on ought to be able to make it out onto the A-Gole. The JIVE client is an NSI Requester, without an NSI provider at the moment as we do not yet manage our own equipment through NSI.

Wouter Huisman (SURFnet): So far, only the dedicated F10-switch is connected, there is no connectivity towards A-Gole or elsewhere yet, apart from the path towards OSO that is being prepared.

Status NSI work and network infrastructure

This was already covered in the updates above.

Deliverable D6.07, Demo international BoD 4Gb/s

The deadline for this deliverable was in December, but we have not yet performed the demonstration. Originally, DANTE and others were hoping to perform a demonstration in December for the EC and had submitted an abstract for this. This demonstration would then also be the demonstration needed to complete the deliverable, Although the abstract did not get selected, they are still planning to perform the test, as the link would also be needed for further deliverables later in the project. But at the moment they are awaiting completion of a requested 4.2Gb/s line, and also as we just heard, the installation of a new router in FUNET. Once these remaining hurdles are cleared, it should be possible to do the demonstration somewhere late in februari. Another option might be to use the “NEXPReS-1” machine that is located in London which has a 10G interface, in case the 4.2Gb/s link to JBO will not become avaialble on time.

The deliverable itself, “Demonstration of international BoD at 4Gb/s”, could possibly also be performed via NSI now that JIVE has a completed a working NSI client. In order to achieve the deliverable as soon as possible, we will attempt both the Metsahovi-JBO path (which is also needed for deliverables in task 3 later on) and the Onsala - JIVE path (which is need for deliverable D6.01 anyway). At OSO, there is still a server from EXPReS that can generate 4Gb/s of data. Richard Hughes-Jones noted that there is hardly any overlap in networking or people resources to create both paths, so doing the one should not delay the other. A seperate telecon will be organised on short notice to discuss what is needed to to enable NSI from OSO to JIVE.

Wouter Huisman asked how we would then set up such an NSI path, as OpenDRAC does not control resources within NORDUnet. JIVE responded that we would use the NEXPReS client to both interact with OpenDRAC at SURFnet, and OpenNSA at NORDUnet, and piece together a complete path. Although NSI defines a tree model and a chain model for forwarding the request, this is not yet implemented in most NSI servers.

Deliverable D6.01, Proof-of-concept BoD client

The good news is that the NEXPReS NSI-client is working, so we have a proof-of-concept BoD client, and are planning to demonstrate it on the OSO-JIVE path discussed in the previous agenda item.

A.O.B.

Richard Hughes-Jones: Can NORDUnet support both NSI and IDCP? Fredrik Pettai: OpenNSA will be NSI based, and we will have a separate testbed for OSCARs using NSI. Richard Hughes-Jones: We have another 10G box in Stockholm that we could use to generate traffic, just would need to be connected again.

Ralp Spencer: when's the next telecon, so that we can keep pushing this work forward? After some discussion, we settled on another telecon in a month, and keeping one another informed in e-mail.

nexpres/nexpres_wp6/telecon_2012_01-11.txt · Last modified: 2012/01/17 15:21 by boven
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 3.0 Unported
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki