Author(s): Richard Schilizzi e-mail: schilizzi@skatelescope.org Affiliation: SKA Project Title: What is e-VLBI? Abstract: The drive to achieve high angular resolution with ever higher sensitivity for astronomy, astrometry and geodesy has taken us through a number of generations of tape recording equipment, into the hard-disk systems now being deployed, and it looks likely to take us into optical-fibre-linked systems within a few years. In this talk I will give a brief overview of the historical development of VLBI before going into the scientific, technical, and operational arguments underpinning the transition into the e-VLBI era.
Session 1: eVLBI Status Reports
Author(s): Alan Whitney e-mail: awhitney@haystack.mit.edu Affiliation: MIT Haystack Observatory
Title: e-VLBI Development at Haystack Observatory Abstract: Haystack Observatory continues an aggressive program of e-VLBI development, particularly with respect to the use of public (shared) high-speed networds for data transfer. Much of 2002 was spent preparing for a Gbps e-VLBI demonstration experiment using antennas at Westford, MA and Greenbelt, MD; this experiment was succcesully conducted using both near-real-time and real-time data transfers to the Mark 4 correlator at Haystack Observatory, though correlation was not done in real time. In early 2003 a dedicated e-VLBI Gigabit-Ethernet wavelength was establisted between Haystack Observatory and MIT Lincoln Laboratory, giving Haystack easy access to the high-speed Abilene network in the U.S. Also in October 2002, preliminary e-VLBI experiments were conducted between Westford, MA and Kashima, Japan; this set of experiments is continuing with increasing data-rate transfers. These experiments use the Mark 5 system at Westford and the K5 system at Kashima; data is transferred in both directions and correlated at both sites. Preparations are now underway to begin e-VLBI transfers from Wettzell, Germany and Kokee Park, Kauaii for routine daily observation of UT1. Haystack Observatory has recently been awarded a 3-year grant the the National Science Foundation for the development of new IP protocols specifically tailored for e-VLBI and similar applications. Author(s): Yasuhiro Koyama , Hiro Osaki, and Tetsuro Kondo e-mail: koyama@crl.go.jp Affiliation: Communications Research Laboratory
Title: Results of test e-VLBI experiments with the K5 VLBI system Abstract: A geodetic VLBI test experiment was performed with K5 VLBI
system from January 31 to February 1, 2003 by using the
Kashima 11-m and Koganei 11-m VLBI stations. The observed
data were processed for software correlation and the site
coordinates of the Koganei station were successfully
estimated from the processed data. K5 VLBI system is
currently under development to realize real-time VLBI
observation and correlation using commonly used shared
network. The system is also capable to record sampled
digital data in the internal hard disks for near real-time
VLBI processing. These capabilities allow to transfer
observed data in real-time if the connecting network is
fast enough, or in near real-time if data buffering is
required. Similar test e-VLBI experiments are planned to
be performed between Westford and Kashima VLBI stations by
using trans-Pacific IP connection. In these test
experiments, data correlation between the different
hardware systems used at two stations will be challenged. Speaker: Yasuhiro Koyama is a Senior Research Scientist at Kashima Space Research Center of Communications Research Laboratory, Japan. He has been involved in the data analysis, software developments, and hardware developments relating to the geodetic Very Long Baseline Interferometry since 1988. Author(s): Hisao Uose, Kazunori Kumagai e-mail: uose.hisao@lab.ntt.co.jp kumagai.kazunori@lab.ntt.co.jp Affiliation: NTT Service Integration Laboratories Title: Network Upgrade of GALAXY project Abstract: GALAXY is a Japanese research project promoted by NTT Laboratories, NAOJ, CRL, and ISAS. NTT has constructed a high speed ATM network dedicated to this project in 1996. Since 2000 we have started to upgrade this network to fully deploy the advance Internet technologies in the field of realtime VLBI. This upgrade enable us collaborate with other research networks. This presentaion describes our present activities and future plans. Author(s): Gerhard Kronschnabl, Reiner Dassing e-mail: kronschnabl@wettzell.ifag.de Affiliation: BKG, Fundamentalstation Wettzell
Title: E-VLBI-activities at the FS Wettzell Abstract: The FS-Wettzell carries out the daily-INTENSIVE observations which were required for the
rapid determination of DUT1. The data volume is roughly 40 GB. So fare the data were
shipped via currier services to the correlator which requires 2-3 days transportation time.
The INTENSIVE time series is a real candidate for E-VLBI. It will reduce the delay due to
data transport strongly. Considering the remote location of Wettzell - apart from the fast
INTERNET links, considering the current high cost for a fast connection, in the next future
the installation of a 34 Gbps-internet connection will be realistic. It will strongly support
the data transmission on start the delay time to only a few hours.
This report give an overview about the activities on the realisation of such a fast link.
First attempts are reported made from the next nodal point at the University Regensburg,
making use of a 155Mbps connection. Author(s): Jouko Ritakari e-mail: Jouko.Ritakari@hut.fi Affiliation: Metsähovi Radio Observatory
Title: eVLBI experiments in Finland Abstract: Metsähovi Radio Observatory has pioneered in capturing VLBI data with
off-the-shelf microcomputers and treating it as normal data that is
stored in normal Linux files. The concept has several advantages over
the traditional thinking where data is stored into "tape-like" special
formats.
During the last year MRO has performed the first intercontinental
1Gbit/s eVLBI experiment with the Kashima radio telescope in
Japan. Since then, experiments have been repeated several times and
the operation of equipment has proven to be extremely robust. These
experiments use direct IF sampling with the Japanese ADS-1000 sampler
currently on loan from CRL and were correlated at Kashima with a
high-speed software correlator.
MRO designed equipment have also been used in the PCEVN project and in
the iGrid eVLBI demonstration.
The MRO data acquisition system has been mass-produced and it is
available off-the-shelf. It is compatible with Mark5P, Mark5A,
Mark5B, ADS-1000, VLBA, and S2 systems. Speaker: Jouko Ritakari received a M.Sc. (economics) from the Helsinki School of
Economics and Business Administration in year 1979 and next year a M.Sc.
(Electronics) from the Helsinki University of Technology. He has 20 years
of experience in designing data communications and fiber optic networks
and 15 years of experience in VLBI. His past projects include designing a
fiber optic network for the Otaniemi campus in 1984 (at the time the
second largest local area network in Europe), designing a fiber-optic
backbone network for the city of Helsinki in 1990 (90 kilometers of fiber
optic cable, 2000 kilometers of fiber, at the time one of the largest
private networks in the world) and connecting the cities of Helsinki and
Vantaa to Internet in 1992 (total of 50000 employees and 150 schools with
70000 pupils). He does hardware design and macroeconomic forecasting, too. Author(s): Steve Parsley, Michael Olberg e-mail: parsley@jive.nl Affiliation: JIVE, Onsala Space Observatory Title: EVN-NREN Proof of Concept Project Abstract: The European VLBI Network (EVN) has access to multiple data sources that can deliver 1Gb/s each and a data processor that can process up to 16 such data streams simultaneously. The EVN can therefore benefit immediately from the multi-gigabit connections that are becoming available in many European National Research and Education Networks (NRENs) and across Europe via GÉANT. Last year SURFnet, the Netherlands NREN, installed a direct fibre-optic link from the Amsterdam Internet Exchange to Dwingeloo, location of JIVE and the EVN Data Processor. This was used successfully at the iGRID conference in Amsterdam to stage the first demonstration of VLBI via international fibre-optic networks. Plans for a larger scale demonstration are now in place. In the next two years up to six European telescopes will be connected directly to the JIVE data processor via NRENs and GÉANT. Trans-Atlantic experiments are also possible via SURFnet and Netherlight. Current concepts for these projects and some of the challenges involved will be discussed. Speaker: Stephen M. Parsley is Head of Technical Operations at the Joint Institute for VLBI in Europe (JIVE) in the Netherlands. His involvement in Radio Astronomy began in 1993 when he joined Penny & Giles Data Systems Ltd. in Wells, UK. There he managed several projects developing and manufacturing data acquisition and recording equipment for the VLBI community. Before this he worked in the UK defence industry developing airborne display and data recording systems for military aircraft. This included holographic head-up displays for General Dynamics F16 and compact, multiplexed video recording and replay equipment for Tornado. Steve is a Chartered Engineer, Member of the Institution of Electrical Engineers and registered with FEANI as a European Engineer.
Session 2: International Network Developments
Author(s): Bernhard Fabianek e-mail: Bernhard.Fabianek@cec.eu.int Affiliation: European Commission
Title: e-Infrastructures - the foundation of the ERA Abstract: Europe has created a pan-European research networking infrastructure – GÉANT – to exploit the emerging developments in telecommunication technology and to support all fields of collaborative research. This infrastructure provides an "equality of opportunity" for all researchers and students across Europe, reaching over 3,100 research and education institutions in over 32 countries through the National Research and Education Networks (NRENs).
As an essential infrastructure that supports all other fields of collaborative research, GÉANT is of paramount importance for the practical realisation of a European Research Area (ERA), of which the European Community's Sixth RTD Framework Programme is a major component.
GÉANT is the leading research network backbone in the world. It not only serves the European continent, but has extensive links in the Gigabit/s range to North America and Japan. Other initiatives ensure the connectivity within Latin-America and the Mediterranean and links between these regions to GÉANT. Activities to link GÉANT with Southern Africa and the Asia Pacific region are in early planning stages.
Thus, the realisation of global research networking is within reach and will enable new ways of research to be carried out. The specific services GÉANT is offering for the VLBI, Grids and advanced communication technologies communities are first examples of the impact that e-Infrastructures will have. In Europe GÉANT is a cornerstone of the European Research ERA (ERA) by providing a world-class advanced communications platform to all researchers and scientists in Europe. Speaker: Bernhard M. Fabianek received a degree in Physics from the Vienna University of Technology for research in the area of semiconductor lasers. He is presently with the Information Society Directorate of the European Commission, Brussels, Belgium, working as a Scientific Officer in the area of Research Infrastructure. His responsibilities include the supervision of the GÉANT project providing the world leading European research network backbone. Author(s): Erik Radius e-mail: erik.radius@surfnet.nl Affiliation: SURFnet bv
Title: NetherLight, the experimental optical internet exchange in Amsterdam Abstract: Since 2001, SURFnet (the service provider for the research and education community in the Netherlands) is deploying novel optical networking techniques at NetherLight, an experimental optical internet exhange for next-generation (multi)gigabit networking.
In this talk, Erik Radius will introduce current and future technologies within NetherLight, the emerging lambda grid at the national, European and transatlantic scale, and recent terabit transport results in collaboration with international partners in Europe, USA and Canada. Speaker: Erik Radius is a manager of Network Services at SURFnet, the national computer network for higher education and research in the Netherlands. He received a M.Sc. in Experimental Physics from the University of Amsterdam and since worked at the University of Twente and at KPN Research on optical network innovations for backbone, metro and access networks. After a short stint in the Internet startup Bredband Benelux, during which he designed the optical part of a Fiber To The Home network infrastructure, he joined SURFnet to work on lambda networking and the closer integration of IP over optical networks. In that role, Erik is responsible for NetherLight, the experimental optical internet exchange in Amsterdam. www.surfnetters.nl/radius/ Author(s): Maxine Brown e-mail: maxine@uic.edu Affiliation: University of Illinois at Chicago/ StarLight
Title: Trans-Light: An Intercontinental LambdaGrid for e-Science Abstract: Trans-Light is a proposed international-scale "LambdaGrid" of electronically and optically switched (Layer 2) circuits and advanced grid services. StarLight in Chicago and SURFnet/NetherLight in Amsterdam, in cooperation with CANARIE and CERN, will be founding members of Trans-Light. At recent NSF Workshops, e-scientists requested high-bandwidth connectivity “…with known and knowable characteristics.” We believe that Trans-Light is the way to support prototypes of the many aggressive e-science applications coming this decade. Trans-Light will provide high-performance capabilities that complement, but do not replace, production international high-performance Internet services. Speaker: Maxine D. Brown is an associate director of the Electronic Visualization Laboratory (EVL) at the University of Illinois at Chicago (UIC), responsible for the funding, documentation, and promotion of its research activities. In 1995 she served as one of several USA technical advisors to the G7 Global Interoperability of Broadband Networks (GIBN) activity, which was a precursor to her being co-principal investigator of the NSF STAR TAP, StarLight and Euro-Link initiatives to provide a persistent infrastructure to facilitate the long-term interconnection and interoperability of advanced international networking. Brown was project manager of the SC'95 Information Architecture/I-WAY/GII Testbed, as well as the iGrid 98, iGrid 2000 and iGrid 2002 high-performance application demonstrations. Along with Tom DeFanti and Bruce McCormick of Texas A&M University, she co-edited the landmark National Science Foundation report, "Visualization in Scientific Computing." Brown has a long history of service to the computer graphics and supercomputing communities, and has been active in both the ACM SIGGRAPH organization and the Supercomputing (SC) conferences. In recognition of her services to the University and the community at large, Brown was a recipient of the 1990 UIC Chancellor's Academic Professional Excellence (CAPE) award, the 1998 ACM SIGGRAPH Outstanding Service Award, and the 2001 UIC Merit Award. Author(s): Cees deLaat e-mail: delaat@science.uva.nl Affiliation: University of Amsterdam Title: High performance Networking for Grid Applications Abstract: The future of networking is to move to an entirely optical
infrastructure. Several leading National Research Networking organizations are creating test-beds to pilot the new paradigm. This talk explores some thoughts about the different usage models of optical networks. Different classes of users are identified. The services, required by the Internet traffic from those different classes of users, are analyzed and a dfferentiated Internet architecture is proposed to minimize the cost per transported packet for the whole architecture. Speaker: Cees de Laat is senior scientific staff member of the Informatics Institute at the University of Amsterdam. de Laat received a PhD in Physics from the University of Delft. He has been active in data acquisition systems, heavy ion experiments and virtual laboratories. Over the past seven years he has been investigating applications for advanced research networks. Current projects include optical networking, lambda switching and provisioning, policy-based networking and Authorization, Authentication and Accounting architecture research. He participates in the European DataGrid project and the Dutch ASCII DAS project. He is responsible for the research on the Lambda switching facility ("NetherLight"), which is currently being built in Amsterdam as a peer to StarLight in Chicago. He implements research projects in the GigaPort Networks area in collaboration with SURFnet. He currently serves as Grid Forum Steering Group member, Area Director for the Peer to Peer area and GGF Liaison towards the IETF. He is co-chair of the IRTF Authentication, Authorization and Accounting Architecture Research group and member of the Internet Research Steering Group (IRSG). http://www.science.uva.nl/~delaat Author(s): Dai Davies e-mail: dai.davies@dante.org.uk Affiliation: DANTE Title: Advanced Networking for Projects Abstract: The cost of international networking has dropped dramatically in the last three years. This has enabled science projects to consider network solutions which a decade ago would have been unimaginable. This raises, however, a number of technical challenges. In particular, whether the current internet technology is the most appropriate solution to meet the requirements of big science projects. In addition, although network liberalisation has been dramatically successful for most of Europe, there are still European locations where liberalisation has yet to have any effect. In a global context, the position is also complicated.
This presentation will review the changes that have taken place, examine the opportunities that these changes have created and give a view of future networking looking up to five years ahead. Speaker: Dai Davies has Degrees in Engineering and Computer Science from the University of Cambridge and nearly thirty years of technical and commercial experience in the telecommunications sector working at BT, Deutsche Bundespost Telekom and the UK Department of Trade and Industry.
After graduation, he worked as a development engineer in BT on a variety of telephony switching and signaling systems. In 1980, he moved to Germany to set up a European technical team working on ISDN implementation. Following his return from Germany in 1983, he moved to a more commercial role in the international division of BT, where he headed up the Marketing and Business Development division, having responsibility for the introduction of the voice VPN service, FeatureNET, as well as working on the business case for acquisition of Tymenet and being responsible for the marketing case for Syncordia, which subsequently became Concert. In 1991, he left BT to run a European project based in the Netherlands that was responsible for building Europe's first pan-European research network. This led to the creation of DANTE.
Dai Davies is now General Manager of DANTE. DANTE was established in 1993 by the University Networks in Europe to organise the provision of International Networking services on their behalf. Its current project, GÉANT, will create a world class pan-European networking facility that complements national developments in Europe. DANTE has organised the procurement in cost sharing of pan-European networking over the last ten years.
His outside interests include reading political biographies, playing the saxophone and travelling. Author(s): T. Charles Yun e-mail: tcyun@internet2.edu Affiliation: Internet2
Title: Update on Internet2 Activities Abstract: [14 Feb 2003]
I propose an overview talk/update on several Internet2 activities that might be of interest to the e-VLBI community including: End-to-End performance project (PiPES); national (American) fiber infrastructure (a.k.a. National Light Rail, FiberCo, etc.); request for interest in short-term, dedicated, test-labmdas (inside the US only at this point).
Session 3: Related Projects
Author(s): Marco de Vos e-mail: devos@astron.nl Affiliation: ASTRON/LOFAR
Title: Networks and processing for LOFAR Abstract: The LOw Frequency ARray (LOFAR) is a high-sensitivity, high-resolution astronomical imaging and detection instrument for low radio frequencies (10-220 MHz). This is a new science window that is of interest for the study of cosmology and astrophysics, as well as for ionospheric studies and space weather applications. Antenna signals are digitised almost immediately after the antenna. This gives LOFAR a unprecedented flexibility, allowing for a wide range of applications. In imaging mode, LOFAR combines the concepts of phased arrays and aperture synthesis imaging. A large number of relatively simple, low cost antennas is employed, distributed over some 100 stations. Phased array beamforming is employed at station level. These data are then transported to a central processing location, where station beams are correlated to form high-resolution synthesis images.
The concept of LOFAR heavily relies on fast datatransport networks and distributed processing power. The total digitized datarate produced by all antennas is roughly 20 Tbps (200 Gbps/station). The station-level processing functions will reduce this to roughly 20 Gbps/station. The paper describes the overall architecture of this networked radio telescope and discusses the major challenges involved. Author(s): Roshene McCool e-mail: rmccool@jb.man.ac.uk Affiliation: Jodrell Bank Observatory, The University of Manchester
Title: e-MERLIN & ALMA - fibre linked radio telescope developments at Jodrell Bak Observatory Abstract: This paper will detail the fibre development work ongoing at The University of Manchester for both the e-MERLIN and ALMA projects. The challenges associated with implementing these two fibre networks will be highlighted. Solutions to overcome these challenges will be proposed, including a description of the optical transmission designs for both telescopes. Speaker: Roshene McCool is the fibre optics engineer at Jodrell Bank. She was awarded an MEng in Electrical & Electronic Engineering, with French, from Nottingham University in 1994. Since graduating Roshene has worked as a project engineer in the telecommunications industry. She joined the fibre optics group at Jodrell Bank in October 2000. In her role as a fibre specialist to the group her responsibilities include the design and development of fibre links involved in IF data transfer for the ALMA project. She also undertakes advisory work for other projects involving fibre at Jodrell Bank, including e-MERLIN and e-VLBI. Roshene is a member of the Institution of Electrical Engineers (IEE). Author(s): Steve Durand, Jonathan D. Romney e-mail: jromney@aoc.nrao.edu Affiliation: National Radio Astronomy Observatory Title: EVLA/NMA: Within and Beyond the 21-km Radius Abstract:
NRAO's Expanded Very Large Array (EVLA) project is being
implemented in two phases. Each involves extremely wide-
bandwidth data transmission over optical fibers, but the
two phases necessarily involve quite different approaches
to the required fiber infrastructure, which make for an
interesting contrast.
Phase 1, formally called the "Ultrasensitive Array",
involves replacing almost all of the existing electronics,
leaving only the mechanical and track infrastructure of
the VLA. The data transmission system being implemented
for Phase 1 uses dedicated optical fibers, currently being
buried at the VLA site. Twelve standard single-mode
fibers will run from each of 72 antenna pads to the
central building. One of these fibers will support the
wideband data transmission system, using a dense wavelength
division multiplexing technique to carry a bandwidth of 96
Gbps (120 Gbps formatted) per antenna. Fibers from the 27
active antenna pads will carry a total bandwidth of 2.6
Tbps. The longest of these fibers will extend the full 21-
km length of each arm.
Phase 2 will add the "New Mexico Array". Eight new
stations will be built, and the electronics of the VLBA Pie
Town and Los Alamos stations will be upgraded, to create a
medium-resolution array, with sensitivity even higher than
Phase 1. All ten NMA stations will lie within the State of
New Mexico. The new antennas will range as far as 265 km
from the VLA site, and will be located so as to facilitate
access to existing fiber trunks installed, primarily, by
rural telephone companies. These trunks include numerous
unused fibers which, it is anticipated, can be leased
economically. The longest fiber run from the VLA is 480 km.
The same 96-Gbps total bandwidth per station will be
supported, with the same underlying sub-band structure.
Signals from up to three NMA stations will be multiplexed
onto a single fiber in the existing trunks. This will limit
the total length of fiber which must be leased or acquired
to about 1240 km. Speaker: Steven Durand holds a Master of Science in Electrical Engineering from New Mexico State University and a Bachelor of Science in Chemistry from Eastern Washington University. He became a member of the technical staff at the National Radio Astronomy Observatory in Socorro, New Mexico, USA in 1999. As the Group Leader for the Data Acquisition Group, he was responsible for maintaining the tape recording systems for the Very Long Baseline Array (VLBA) including developing QA procedures and troubleshooting techniques. He presented over 50 classes on recorder maintenance techniques. As Electronics Task Leader for the Expanded Very Large Array (EVLA) Fiber Optic Group, he designed the fiber optic system to be used for the IF, LO and MCB systems. Durand currently holds the position of Electronics Division Head where he is directing the development of the electronics for the EVLA. At the same time, he is responsible for the maintenance of the electronics in the Very Large Array (VLA) and the VLBA. He has published a number of papers on optical digital transmission systems, photovoltaic power systems and hybrid power systems.
Session 4: eVLBI Technology
Author(s): Ralph Spencer, Steve Parsley and Richard Hughes-Jones e-mail: res@jb.man.ac.uk Affiliation: University of Machester, JIVE Title: The resilience of e-VLBI data to packet loss Type of presentation: Abstract: Demonstrations of internet VLBI data transfer were made last
year at iGRID2002 and at the EU's FP6 Launch. Data rates of
> 500 Mps were achieved with some packet loss . This paper
summarises these results and shows that packet loss does not
result in loss of data frames unless loss rates are high,
though de-correlation will occur. Speaker: Dr Ralph Spencer is Reader in Radio Astronomy at the University of Manchester. He has worked on the development of interferometers for radio astronomy since the 1970’s, leading work on the development of phase stable radio linked interferometers with 100 km baselines. This work led to the development of the MERLIN array of 7 telescopes now operated as a National Facility by the University of Manchester for PPARC. His interest in VLBI started in the late 1970’s and was responsible for running VLBI operations at Jodrell Bank Observatory until the mid 1990’s, when the National Facility took over. European development of the 1 Gbps MkIV VLBI system was led by him until 2001, and this is now in routine operation world wide at data rates up to 256 Mbps. His astronomical research interests include studies of microquasars. These are radio emitting X-ray binaries in our own galaxy which emit jets of matter at relativistic velocities. They act as laboratories for the study of the formation of jets in general, a topic which is great relevance to the study of quasars and radio galaxies in general. Ralph is a Fellow of the Royal Astronomical Society and a Member of the New York Academy of Sciences. Author(s): Richard Hughes-Jones, Ralph Spencer and Steve Parsley e-mail: R.Hughes-Jones@man.ac.uk Affiliation: University of Manchester, JIVE
Title: High Data Rate Transmission in High Resolution Radio Astronomy using International Academic Networks – vlbiGRID Abstract: Recent developments in Very Long Baseline Interferometry (VLBI) radio astronomy are aimed at improving reliability and reducing the cost of operations by moving from a custom made tape based system to using exchangeable disks in PCs, and then linking the Telescopes with the Correlator using the internet. This paper describes the work done to investigate the possibility of transmitting VLBI data over the production academic networks. VLBI data was sent from a server at Manchester, over the Net North West Metropolitan Network, MAN, and onto the UK Academic Network, SuperJANET4. The GÉANT backbone provided the international connectivity between the UK and SURFnet, the academic network in the Netherlands, and this in turn was connected to servers in Amsterdam and Dwingeloo that acted as data sinks.
These tests were demonstrated at the iGrid2002 exhibition held in Amsterdam in September 2002 and at the European Research 2002 exhibition in Brussels in November 2002 which marked the launch of the European Union Framework 6 Initiative. Collaboration with UKERNA, SURFnet and Dante together with the availability of high data rate international links gave us the opportunity to try transferring the data via the production internet.
The tests were successful achieving data rates from Manchester of 500 Mbps over the production network. This was the first demonstration of fibre-optic link connected international VLBI.
Discussion and details of the UDP/IP transfer mechanisms used are presented together with network performance measurements made during the tests. The effect of operating with different Quality of Service traffic on the backbone is also presented. The VLBI Demonstration confirmed that the end hosts must have sufficient power in both compute cycles and input/output capability; the requirements for operation at Gigabit speeds will be shown. Some recent measurements made on modern server-quality PC motherboards and Network Interface Cards are also presented. Speaker: Richard Hughes-Jones leads development of the Grid and Trigger/Data Acquisition systems in the Particle Physics group at Manchester University. He has a PhD in Particle Physics and has worked on Data Acquisition and Network projects for over 20 years, including evaluating and field-testing OSI transport products. His interests are in the areas of realtime computing and networking within the context of the ATLAS experiment in the LHC programme including the performance, network management and modelling of Gigabit Ethernet components. Within the Global Gird Forum he is co-chair of the Network Measurements Working Group. He is secretary of the Particle Physics Network Coordinating Group which has the remit to support networking for PPARC funded researchers. Within the UK GridPP project he is deputy leader of the network workgroup and is active member of the EU DataGrid networking workpackage (WP7). He is also a PI in the EU DataTAG project and involved in the UK MB-NG project to investigate QoS and various traffic engineering techniques. He is currently investigating the performance of transport protocols over LANs, MANs and WANs. Author(s): Hans Hinteregger e-mail: hhinteregger@haystack.mit.edu Affiliation: MIT Haystack Observatory
Title: Capability and Cost of COTS Networking Components Abstract: In this talk we will emphasize first the ease with which dedicated 'dark' fiber can today be 'lit' at one or more gigE-bandwidth wavelengths with low-cost switches, optical transceivers with up to 100 Km reach, and passive mux/demux's. The chief expense for dedicated links today is that of leasing a fiber pair, especially in a city like Wash. DC. But, at least in the US, the intercity cable overbuild of the telecom boom has now, in the bust, made in-the-ground dark fiber pairs buyable for <$1000/km. Hence there are now business interests in COTS networking components,with corresponding opportunities for e-VLBI. We will outline the fairly clear and certain road from gigE in SFP to 10gigE in XFP commodity pluggable packages. The timing of cost/bandwidth crossover is less so -- due to current uncertainty in the telecom business climate. A 2006 budget for LOFAR networking requirements, assumed here to be built from scratch in the WA desert where the dominant cable laying costs are minimized, is considered in some detail. The cost of fibers and the cost lighting them (one-way) to 200 Gb/s per station is judged to be affordable. A further projection to SKA's COFON in 2009 will be made. Denser CWDM and/or cheaper DWDM may be developed and mature by that time. Author(s): Ari Mujunen e-mail: Ari.Mujunen@hut.fi Affiliation: Metsähovi Radio Observatory
Title: Applying off-the-shelf technologies in eVLBI Abstract: This is an overview of available COTS disk, tape,
computer, and networking technologies illustrated
with evaluation results achieved in actual
Metsähovi Radio Observatory tests. Current
technology limitations, their impact on system
design and system scalability, and near-future
COTS trends are being discussed. A strategy to
overcome technology limitations when developing an
uninterrupted Gbit/s eVLBI system is presented. Speaker: Ari Mujunen is a software engineer at Metsahovi Radio Observatory. He has
a M.Sc. in EE from Helsinki University of Technology with software
engineering and production as the major subject. Prior to joining MRO in
1992 he worked for commercial software houses creating software
production toolsets for both CAD and business database applications. At
MRO, he has written software and created electronic designs for the
in-house telescope control and single-dish data acquisition system. For
the 1998 STS-91 space shuttle AMS mission MRO was responsible for the
ground segment of AMS HRDL (High-Rate Data Link) for which Ari Mujunen
wrote the autonomous Linux-based data acquisition software package.
He soldered together his first home computer in 1978 (a derivative of the
RCA COSMAC 1802 ELF II) and he has ever since wished that software
development tools were more powerful and productive.
Author(s): Hisao Uose e-mail: uose.hisao@lab.ntt.co.jp Affiliation: NTT Service Integration Laboratories Title: VLBI data transmission system using multiple IP streams Abstract: We have developed an IP-based data transmission system which can directly replace K4 (ID1) VLBI data recorder. It employs multiple TCP data steams and individual buffers to cope with variable usable bandwidth we encouter when we use shared academic networks. This paper describes the outline of the system and preliminary testing results. Author(s): David Lapsley e-mail: dlapsley@haystack.mit.edu Affiliation: MIT Haystack Observatory
Title: Development of New IP Protocols for e-VLBI Abstract: One of the key challenges to be met in implementing eVLBI is how to transport large amounts of VLBI data across public infrastructure at low cost without impacting other users' traffic. In this presentation, we will discuss the approach being taken at the MIT Haystack laboratory. Our approach is based on two fundamental scenarios. On the one hand, we consider data that is transported across a network that provides differentiated service. This scenario does not require any special intelligence at the end systems as it relies on the network to ensure that VLBI traffic does not impact other users' traffic. On the other hand, we consider the transport of VLBI data across a network that does NOT provide differentiated services. In this case, we propose an experiment guided adaptive transport protocol. The characteristics of the dynamic protocol will be configurable through the use of experimental "profiles". The transport protocol itself will adjust the rate of VLBI data transmission to match the available network resources while maintaining the quality of service required to support VLBI. We provide an architectural framework for using these two transport mechanisms to support eVLBI, we discuss the functionality of the end systems that implement the adaptive transport protocol and we discuss candidate transport protocols for supporting it. Author(s): Ian Avruch, Sergei Pogrebenko e-mail: avruch@jive.nl Affiliation: JIVE Title: A Distributed Software Correlator Abstract: A number of groups around the world are implementing the functions of
VLBI correlators on general-purpose computers, in software. The
tradeoff is decreased capacity versus lower cost and greater
flexibility. For the near future, dedicated hardware will remain the
only option; for example, the EVN correlator at JIVE is equivalent to
several thousand current-generation desk-top PCs requiring I/O rates of
1Gbps. However, the trends in I/O and computational bandwidths imply a
distributed software correlator will soon be a possible and perhaps
preferable solution. Speaker: Ian Avruch is a Correlator Support Scientist at JIVE.
Posters
Author(s): Kimura Moritaka, Junichi Nakajima, Tetsuro Kondo e-mail: mkimura@crl.go.jp Affiliation: CRL, Kashima Space Research Center, JAPAN
Title: Multi baseline Grid Software Correlator Abstract: Software VLBI correlation is regarded as a solution for next generation VLBI.
With a flexibility of the software correlation programming, appropriate
scientific correlations by scientists are possible as well as the
post processing.
As the first experiment to handle Gbps VLBI data, multi baseline
Grid correlator have been developing at CRL.
The performance of software correlation adopted multi CPUs, SIMD architectures
and Grid computing technology has nearly reached hardware correlator performance.
