Table of Contents
IMPORTANT NOTICE - JIVE WIKI MOVING SOON
Over the next few weeks this page will move to . Watch for email with information about new user accounts, and consider making edits at the new site. Questions? Email firstname.lastname@example.org.
FABRIC: Business Meeting
PSNC (Polish Supercomputing and Networking Center)
25 September 2006
- Paul Burgess (JBO)
- Richard Hughes-Jones (JBO)
- Huib van Langevelde (JIVE)
- Ruud Oerlemans (JIVE)
- Arpad Szomoru (JIVE)
- Charles Yun (JIVE)
- Roel Gloudemans (ASTRON)
- Ari Mujunen (MRO)
- Helge Rottmann (MPIfR)
- John Conway (OSO)
9:00 - 9:15 Welcome & logistics; Meyer 9:15 - 9:30 Goal of the meeting; van Langevelde
9:30 - 10:00 State of EXPReS, money flow, reporting; Yun 10:00 - 10:30 Spending, hires, hours spent; van Langevelde/Yun 10:30 - 11:00 Coffee break 11:00 - 11:45 Progress, Gantt charts, milestones; van Langevelde
Active work-packages: progress, bottlenecks, decisions
11:45 - 12:30 WP 1.1 Data Acquisition; Mujunen 12:30 - 14:00 Lunch & tour of the facility 14:00 - 14:30 WP 1.2.1 Broadband protocols; Hughes-Jones 14:30 - 15:00 WP 1.2.2 Broadband interfaces; Burgess 15:00 - 15:30 WP1.2.3 Broadband tests; Conway 15:30 - 16:00 Tea break 16:00 - 16:20 WP 1.2.4 LOFAR; Maat 16:20 - 17:00 WP 2.1 Grid resource allocation; Okoñ/Lawenda 17.00 - 17.20 WP 2.1.3 Networking related topics; Garstka 17:20 - 18:00 WP 2.2 Software correlation; Oerlemans/van Langevelde 18:00 - 18:15 Wrap up, next meeting; van Langevelde
9:00 - 9:15 Welcome & logistics; Meyer
Introduction to the day, building and city. General logistics and background.
9:15 - 9:30 Goal of the meeting; van Langevelde
van Langevelde outlines agenda, covers format and pace.
9:30 - 10:00 State of EXPReS, money flow, reporting; Yun
Yun reviews the current state of the project. The current status of the consortium agreement is discussed which has delayed the money flow. Standards for future reporting and audits are discussed, as well as the conditions for dealing with depreciation.
10:00 - 10:30 Spending, hires, hours spent; van Langevelde/Yun
11:00 - 11:45 Progress, Gantt charts, milestones; van Langevelde
van Langevelde presented accounting/financial reporting information collected from the participants of FABRIC. He identified good examples of reporting and some questions that will need to be added for future reports.
There continue to be questions regarding how spending needs to be identified and tracked. The questions are the logical and important ones, and will be answered directly by project manager (when they can be answered), by the EC science officer (as applicable) or by time as the first full reporting cycle is completed (a less desirable path). Questions include items such as depreciation, audit schedules and frequency of reporting, inconsistencies in timeline.
There is also a note that spending should be identified by activity (e.g., separate SA1 and JRA1) not by location (all JIVE activities in one item).
Overall the progress in the first 6 months is satisfactory, taking into account the delays in finding the manpower necessary. Some places are underspending because of the failed manpower acquisition.
van Langevelde covers deliverables and status by month. Reminds people that the wiki should be posted to the web site.
11:45 -12:30 WP 1.1 Data Acquisition; Mujunen
Mujunen provided a summary of current state of data acquisition systems looking at options for future work and demos. Options include using old hardware, modifying existing hardware and developing new hardware. He identified and discussed some of the presentations at Haystack Observatory (5th Intl eVLBI Workshop) from last week. Further went into details about the iBOB (interface/infiniband break out board) in terms of chips, bandwidth and future options. Also spoke of iBOB compatibility.
Discussion of how to modify process/hardware for demo and how this is functionally different than planning for operational use of the “system” through the future. This surrounded the discussion of iBOBs and buffers between locations. Options were many and consensus leaned towards a iBOB solution.
There was healthy discussion on what the consequences were and the issues involving purchase (order deadlines) as well as cost for the iBOBs.
14:00 - 14:30 WP 1.2.1 Broadband protocols; Hughes-Jones
Hughes-Jones introduces “Protocols Document 1” and how this has been useful to gather consensus in the astronomy community as well as how it is useful for external communities to understand VLBI networking needs through an explicit document. “Protocols Document 2” is now under way. Mentions individuals who are working on it, different protocols under discussion and the conversations from the Haystack meeting.
Presented early results from some network tests using “tcpdelay.” Additional information about the way in which iBOB functions over the (limited) network.
Quick overview of the networking organizations (DANTE/GEANT, fiber, NRENs, placing boxes at pops) and technologies used. Outlining how VLBI community can take advantage of technologies available and what they want to do in terms of cooperation with backbone network and POPs. Relevant items to VLBI include: protocol/UDP, long duration flows (details on slide).
14:30 - 15:00 WP 1.2.2 Broadband interfaces; Burgess
Updates on interfaces and interconnects (station → iBOB → network/disk). Reporting good progress and happiness with early tests. Also notified on the retirement of tape systems.
Connection with/to Onsala is currently at 10 Gbps, hoping for long term 40 Gbps connection, but funding is current roadblock. Lightpaths are a discussion point, but no clear current answers.
Observation that Ethernet/IP are long term standards and should be worked towards directly versus more transient protocols/standards (e.g. variants of PCI). Should aim for long term, scalable standards.
Suggestion for EVN order for 50 iBOBs. This is the production volume break even point for one time costs.
15:00 - 15:30 WP1.2.3 Broadband tests; Conway
Testing for the demo is in process, date selected. Outline of the demo is still a bit vague, but they do believe that they can showcase current activities. Discussion on how to get each of the sites prepared for the demo. John is asking all participants to start testing and contacting JIVE for connectivity immediately.
Large volume of discussion of how to move forward in terms of activities. Presenter acknowledged a more prescriptive and loose approach. Desire for FPGA expertise as well as concerns on up-scoping contract. Some questions about the distribution of work between partners.
The second question is if they should use the same iBOB platform for high speed link and distributed correlation?
16:00 - 16:20 WP 1.2.4 LOFAR; Maat
Have made a variety of software and hardware decisions on protocol, use of lightpaths and locations. Details of hardware specifications are in the slides. Note that decisions were made some time ago and were based on cost effectiveness (e.g., single mode fiber) at that time. LOFAR testing 1 and 10 Gbps Ethernet. Decision to use COTS (common off the shelf) hardware does have some problems.
Future work includes data transfer rates at 2.1 Gbps as lightpaths are not available at all locations.
16:20 - 17:00 WP 2.1 Grid resource allocation; Okoñ
A nice introduction of the way PSNC sees the eVLBI process. A bit of discussion on when the data is “unix friendly” and more discussion is needed to be sure PSNC understands the Mk5 and Mk4 format issues. This is particularly because PSNC is moving forward with assumptions on the way that the telescopes will send their GRID environment software. There is also the system design slide which should also be discussed.
The short term goals and long term goals seem well stated to obtain and clarify some of the points of discussion during his talk.
17.00 - 17.20 WP 2.1.3 Networking related topics; Garstka
Talking about the ways in which network elements will effect the grid software project. There is the concept of the grid gateway. Also discussion of QOS (quality of service) on the network. Not in traditional networking reservation terms, but in terms of the metrics necessary to assure/support good quality connections.
Survey of a variety of tools/technologies: perfsonar and clusterix. Question followed asking if the listed tools are going to integrate with Globus and how this will happen (interfaces). Answer is that there is an interface planned, but this is still early in the investigation.
17:20 - 18:00 WP 2.2 Software correlation; Oerlemans
Oerlemans' presentation covers assumptions, explanations and decisions while developing the distributed software correlator. There is good explanation of the current situation.
Healthy interaction with PSNC coders helping clarify VLBI process from astronomer's point of view and correlation point of view.
18:00 - 18:15 Wrap up, next meeting; van Langevelde
Action Items / Decisions
- get an iBob and begin to work with it
- publish design document w/r/t iBob
- continue to think about Onsala manpower
- demo: quickly start doing dry tests
- testbed/pilot installation of distributed correlator
- needs to meet the iBob acquisition
- need to identify next meeting location
- confirm May 8 as next meeting time date