Transcript Poster: VLBI_UDP - University of Manchester

VLBI_UDP
Simon Casey*, Richard-Hughes Jones#, Ralph Spencer*, Matthew Strong*
* Jodrell Bank Observatory, The University of Manchester; # HEP Group, The University of Manchester
Introduction
Results
VLBI_UDP is an application being developed to transfer VLBI data
using the UDP protocol. Initially developed by Richard HughesJones for continuous streaming of data, features are being added to
allow streaming of VLBI data from the data acquisition devices
PCEVN and Mark5A, as well as the ability to selectively lose UDP
packets. The architecture of VLBI_UDP can be seen in Fig. 1.
Fig. 3 shows a non-stop 24 hour run between Jodrell Bank and
Manchester. Each point represents the average received wire rate
over a 30 second sample, and demonstrates the long term stability
of the application.
00:00
23:00
22:00
21:00
20:00
19:00
18:00
17:00
16:00
15:00
14:00
13:00
12:00
11:00
10:00
09:00
08:00
995.5
995
07:00
Memory
06:00
Disk
05:00
Memory
04:00
Disk
996.5
996
03:00
Control thread
TCP Control
02:00
Control thread
998
997.5
997
01:00
vlbi_recv
999
998.5
00:00
vlbi_send
Wire rate (Mbit/s) s
Vlbi_control
24 Hour flow Jodrell Bank - Manchester
1000
999.5
Tim e of day
Input thread
Output thread
Ring buffer
Ring buffer
UDP Data
Send thread
Receive thread
Fig 3. 24 Hour flow
Fig. 4 shows the GÉANT2 core network, and the telescopes with
high-speed connections used as the test bed. In December 2006, a
3 station e-VLBI experiment was emulated by simultaneously
transmitting data from 3 locations into PCs at JIVE. The results of
this can be seen in the 3 plots of Fig. 5.
Fig 1. VLBI_UDP architecture
Architecture
Gbit link
Chalmers
University
of
Technology,
Gothenburg
Metsähovi
At the sender, the input thread either reads data from a file or
generates random data and places these in to the ring buffer. The
output thread removes packet size chunks of data from the ring
buffer and encapsulates the data part of a UDP container. The
header of this container is then filled with a sequence number which
increments by 1 for each packet sent. This packet is given to the IP
stack which places it on the network.
Onsala
Sweden
Jodrell Bank
UK
Gbit link
At the receiver, the receive thread places incoming packets directly
into the next position in the ring buffer. The sequence number is
read from the header, and this reveals whether the packet is at the
correct position in the buffer. If the sequence number increment is
more or less than 1, then the packet is moved forwards or
backwards in the buffer, as indicated in Fig 2.
Torun
Poland
Dedicated
Gbit link
Dwingeloo
DWDM link
Medicina
Italy
Fig 4. GÉANT2 network & telescope connections
0.8
600
0.6
400
0.4
200
0.2
0
0
2000
4000
6000
8000
10000
12000
0
14000
Time during the transfer s
1
800
0.8
600
0.6
400
0.4
200
0.2
0
0
2000
4000
6000
8000
10000
12000
% Packet loss
1000
g
Manchester - JIVE, NL (Packet switched)
Wire Rate Mbit/s
Packet 3 missing, move packet 4
forwards
Packet 3 out of order, move packet
3 backwards
g
1
800
% Packet loss
Wire Rate Mbit/s
Manchester - JIVE, NL (UKLight)
1000
0
14000
Time during the transfer s
A further mode of operation has been added whereby data are read
from a file and filtered through a packet-dropping function which
selectively drops packets according to the selected algorithm. The
data are then written out to a file without being sent over the
network. This is explained in more detail in the corresponding paper
‘Investigating the effects of missing data upon VLBI correlation using
the VLBI_UDP application’.
0.8
600
0.6
400
0.4
200
0.2
0
0
2000
4000
6000
8000
10000
12000
g
1
800
% Packet loss
The output thread removes blocks from the ring buffer and either
discards or writes them out to a file, as specified by the user.
Bologna - JIVE, NL (Packet switched)
1000
Wire Rate Mbit/s
Fig 2. Simulation of packets being placed into ring buffer
0
14000
Time during the transfer s
Fig 5. Simultaneous UDP flows into JIVE
The absence of packet loss, clearly show the superior performance
of the UKLight lightpath when compared with the packet switched
production network.
This work was performed in collaboration with the EXPReS project, EC FP6 contract number 026642