Radio Astronomy VLBI Data Transmission

During iGrid2002, the link to the correlator at Dwingeloo became operational, and we believe that this was the first time that VLBI data has been successfully recorded at Radio Telescopes in different countries and transferred over the production internet to the correlator.

Information on Radio Astronomy and current VLBI techniques was given in a Web presentation .

This screen shot of the demonstration shows The Progress Monitor frame which is a set of coloured boxes each representing a 1.8 Gbyte data transfer and the VLBI Transmission Statistics frame that shows the throughput, packet loss and number of out of order packets for each 1.8 Gbyte transmission.

A Technical presentation gave details of the techniques used in the demonstration and some performace tests made on the networks.

"High Data Rate Transmission in High Resolution Radio Astronomy - vlbiGRID" is a paper describing the demonstration in detail that has been submitted to a special edition of FGCS, the Future Generation of Computer Systems Journal.

Some of the Measurements made during iGrid2002

The plot of the traffic levels from the SuperJANET4 access router at Manchester for the Net North West MAN during the iGrid2002 meeting is shown below. The normal diurnal traffic levels for the Net North West MAN vary between 70 and 300 Mbi/s into the MAN (solid graph) and between 200 and 400 Mbit/s out of the MAN (line graph). The 500 Mbit/s VLBI traffic for iGrid2002 is visible as the sharp spikes, which occur while the demonstration was in progress and the VLBI data takes the outgoing traffic level to 650 to 700 Mbit/s, or 65 to 70 % of the access links.

Graphs of the user and "wire" data transfer ratesfor a series of 1.8 Gbyte data transfers are shown below. These rates were measured by the programs running the demonstration. Each transfer sent 1.24M packets, each containing1452 bytes of user data. The red curve shows the "wire" transfer rate which inlcudes all the networking and framing headers. The 515Mbit/s rate was determined by selecting the time between transmitting packets to be 20 µs. Normally during the demonstration, about 40 - 50 packets were lost per transfer. The lower plot shows the corresponding number of packets lost and the number that arrived out of order.

Study of the packets that arrived with their sequence numbers "out of order" showed that these packets had been overtaken by the previous packet that had been sent. This effect has not been seen on tests made on the UK academic network SuperJANET4, nor on SURFnet. As no multiple trunk links have been deployed in the networks that were used, this re-ordering effect is believed to be due to the use of parallel forwarding engines in the Juniper M150 routers used in Geant.

The Link to JIVE

While iGrid2002 was in progress the link provided by SURFnet between Amsterdam and Dwingeloo became operational. Data from Manchester were sent by ftp to Dwingeloo, received via a Gigabit Ethernet connection and copied onto disks in a PCEVN system. The data were then transferred to the correlator via a standard VLBI station unit originally designed for use with the MkIV magnetic tape playback machines. Data from the Westerbork and Jodrell Bank telescopes were then correlated (one set was on tape, the other on the disks described above). The correlator cross multiplies the signals (in phase and quadrature) for a range of relative time delays, averages for 2 seconds and produces plots of the correlation coefficient. The plots show the amplitude as well as real and imaginary parts of the correlation coefficient. The peak at a relative delay of 33 clock periods shows that the fringes were clearly detected from DA193 and that the data were of good quality.

Richard Hughes-Jones 20 Nov 02