connect. communicate. collaborate support by gÉant2 and nrens for evlbi in europe john chevers,...
TRANSCRIPT
Connect. Communicate. Collaborate
Support by GÉANT2 and NRENs for eVLBI in Europe
John Chevers, DANTEIn collaboration with:
Toby Rodwell (DANTE)
Steve Parsley, Arpad Szomoru (Joint Institute for VLBI in Europe)
Paul Burgess, Simon Casey, Richard Hughes-Jones, Adam Mathews, Stephen O’Toole, Ralph Spencer (University of Manchester)
TNC2005 Poznan, Poland 8th June 2005
Connect. Communicate. CollaborateUser Support in GÉANT2
• GÉANT2 has 3 million+ users in 34 countries in Europe and beyond• How to support them?• Pan-European projects with ‘demanding’ requirements
– International user groups– High bandwidth requirement?– Specialised service requirement?
• Aim to introduce new levels of accountability, transparency and service• Custom solutions for the particular user group• To manage the lifecycle of the networking aspects of each user project• Customer focused
Connect. Communicate. Collaborate
User Support Mechanism for GÉANT2
• Initial Contact– Establish eligibility of user application– Inform all NRENs involved– Assign point of contact
• User Questionnaire– Requirement gathering
• Evaluation– Consultation between DANTE and NRENs– Define stakeholders
Connect. Communicate. Collaborate
User Support Mechanism for GÉANT2 (II)
• Planning and Implementation– Action plan– Design of OPN– Risk Register
• Continued Support– A service charter for the user– Change management
QUALITY OF CUSTOMER SERVICE IS VITAL!
Connect. Communicate. CollaborateNetworking for eVLBI in Europe
• European eVLBI Network: collaboration of radio observatories in Europe, Asia and South Africa
• Presently 5 telescopes connected: Torun, Onsala, Cambridge, Westerbork, Jodrell Bank
• Many more to be connected soon• Large data flows, monitoring required• A demanding user!
Connect. Communicate. Collaborate
The future of the distant past:using networks for VLBI
• Connecting observatories to create a telescope as big as Europe!
• Continental-scale baselines produce the most detailed images ever of the most distant objects!
• Improving the resolution of observations to see further back in time
• Eliminating out-moded tape technology and speeding up results
• Harnessing the power of many radio telescopes in real time –only possible with GÉANT2
• Allowing new ‘Window of Opportunity’ experiments: dynamic scheduling
• Distributing results to users across the world
Connect. Communicate. CollaborateGÉANT2/NREN eVLBI Network Connect. Communicate. Collaborate
se1 - Juniper
uk1 - Juniper
nl1 - Juniper
cz1 - Juniper
pl1 - Juniper
de1 - Juniper
it1 - Juniper
de2 -Juniper
STM-64 (10Gbps)
ws4.it Dell
geant-gw3lond-scr3
man'-bar
Cisco Cisco
warr-scr
Cisco
read-scr3
Milan GSR
BolognaGSR
Bologna
GARR bwctl Mk5 MB
STM-16 (2.5Gbps)
Gigabit Ethernet
ManCompCen
Transmode
3ComCR1
AR5
JIVE, NL
JODRELL BANK,
UK
Mk5s
Mk5s
BOLOGNA, IT
ws4.ukDell
Mk5 Torun
oso-gw
chalmers2
goteborg2
boras2jonkoping1stockholm1stockholm2
se-kth
Mk5
Cisco
ONSALA, SE
TORUN, PL
Unknown
Netherlight
Connect. Communicate. CollaborateGÉANT2 User Support for eVLBI
• The EVN-NREN forum: the communities meet face-to-face!– Planning– Problem solving– Discussion of results– Increasing understanding
• GÉANT Weathermap and Taxametro tools• Counters on GÉANT links• Diagnostic tools (BWCTL) in GÉANT and NREN network nodes• PERT account created –letting SMEs loose on the problem!• Also use of specialised tools eg UDPmon
Connect. Communicate. CollaborateProgress in eVLBI 2004-2005
• 01/04 First disc-buffered eVLBI image • 04/04 Three-telescope eVLBI session• 06/04 eVLBI network stress tests (iPerf)• 09/04 4 telescope eVLBI• 09/04 First real-science eVLBI • 12/04 Fringes at 256 Mbps JB-Tr• 02/05 Network data transfer tests (BWCTL) between 6 eVLBI sites• 03/05 eVLBI science session involving 6 telescopes: Wb, Ar, JB, Cm, On, Tr• 03/05 EXPReS Proposal:
– ‘Production’ eVLBI– 16 telescopes – High Data Rates– Global eVLBI
• 06/05 Connection of Medicina observatory: brings total to 6 telescopes in 5 countries across Europe
Connect. Communicate. Collaborate
4th Year MPhys ProjectAdam Mathews Steve O’Toole
Throughput vs packet spacing Manchester: 2.0 GHz Xeon Dwingeloo: 1.2 GHz PIII Near wire rate, 950 Mbps Packet loss CPU Kernel Load Receiver
UDP Throughput Nov 2003Manchester-SuperJANET4-GEANT-Surfnet-Dwingeloo
Gnt5-DwMk5 11Nov03-1472 bytes
0
2
4
6
8
10
12
0 5 10 15 20 25 30 35 40Spacing between frames us
% P
acket
loss
Gnt5-DwMk5
DwMk5-Gnt5
Gnt5-DwMk5 11Nov03/DwMk5-Gnt5 13Nov03-1472bytes
0
200
400
600
800
1000
1200
0 5 10 15 20 25 30 35 40Spacing between frames us
Recv W
ire r
ate
Mbits/s
Gnt5-DwMk5
DwMk5-Gnt5
Gnt5-DwMk5 11Nov03 1472 bytes
020406080
100
0 5 10 15 20 25 30 35 40Spacing between frames us
% K
erne
l R
ecei
ver
Effect on local traffic
Connect. Communicate. Collaborate 1472 byte Packets man -> JIVE Relative 1-way delay
Packet Loss & 1-way Delay
1472 bytes w12 Gnt5-DwMk5 21Oct03
0
2000
4000
6000
8000
10000
12000
2000 2100 2200 2300 2400 2500 2600 2700 2800 2900 3000Packet No.
1-w
ay d
elay
us
1472 bytes w12 Gnt5-DwMk5 21Oct03
0
2000
4000
6000
8000
10000
12000
0 1000 2000 3000 4000 5000Packet No.
1-w
ay d
elay
us
Connect. Communicate. Collaborate
Measure the time between lost packets in the time series of packets sent.
Lost 1410 in 0.6s Is it a Poisson process? Assume Poisson is stationary
λ(t) = λ Use Prob. Density Function:
P(t) = λ e-λt
Mean λ = 2360 / s[426 µs]
Plot log: slope -0.0028expect -0.0024
Could be additional process involved
Packet Loss DistributionIs it Poisson?
packet loss distribution 12b bin=12us
0
10
20
30
40
50
60
70
80
12 72 132
192
252
312
372
432
492
552
612
672
732
792
852
912
972
Time between lost frames (us)
Num
ber
in B
in
Measured
Poisson
packet loss distribution 12b
y = 41.832e-0.0028x
y = 39.762e-0.0024x
1
10
100
0 500 1000 1500 2000
Time between frames (us)
Num
ber
in B
in
Connect. Communicate. Collaborate
eVLBI Data Transfers over GÉANT:Long term tests of Bandwidth
Jodrell-UKERNA-GÉANT-SURFnet-JIVE
1472 byte UDP packets, Dual Xeon (Manchester) to Mark5A (JIVE)
550
600
650
700
750
800
850
900
950
01/11/2004 12:00 09/11/2004 12:00Time of day
Wir
e r
ate
(M
b/s
)
12us
18us
20us
Connect. Communicate. Collaborate
eVLBI Data Transfers over GÉANT: Long term tests of Packet Loss
Jodrell-UKERNA-GÉANT-SURFnet-JIVE
• 6 runs out of 747 showed any loss• 1097 packets lost out of 81 million sent – 1.3x10-3%
1472 byte UDP packets, Dual Xeon (Manchester) to Mark5A (JIVE)
0
0.2
0.4
0.6
0.8
1
1.2
01/11/2004 12:00 09/11/2004 12:00
Time of day
% L
os
t 12us18us20us
Connect. Communicate. Collaborate
UDP Tests Jan 2005 Simon Casey (PhD project)
Between JBO and JIVE in Dwingeloo, using production networkPeriod of high packet loss (3%):
Connect. Communicate. Collaborate
GEANT – UKLight Comparison May 2005 Simon Casey (PhD project)
Between JBO and JIVE in DwingelooUsing production network Using UKLight network
gig6-jivegig1_int_1May05
050000
100000150000200000250000300000350000
0 10 20 30 40Spacing between frames us
Pkt
s re
-ord
ered
50 bytes
100 bytes 200 bytes
400 bytes 600 bytes
800 bytes 1000 bytes
1200 bytes 1400 bytes
1472 bytes
gig6-jivegig1_int_1May05
0
20
40
60
80
100
0 10 20 30 40Spacing between frames us
% P
acke
t lo
ss
50 bytes
100 bytes 200 bytes
400 bytes 600 bytes
800 bytes 1000 bytes
1200 bytes 1400 bytes
1472 bytes
gig6-jivegig1_int_1May05
0100200300400500600700800900
1000
0 10 20 30 40Spacing between frames us
Recv W
ire r
ate
Mbit/s
50 bytes
100 bytes
200 bytes
400 bytes
600 bytes
800 bytes
1000 bytes
1200 bytes
1400 bytes
1472 bytes
gig7-jiveg1_UKL_31May05
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40Spacing between frames us
Pac
ket
re-o
rder
ed
50 bytes
100 bytes 200 bytes
400 bytes 600 bytes
800 bytes 1000 bytes
1200 bytes 1400 bytes
1472 bytes
gig7-jiveg1_UKL_31May05
0100200300400500600700800900
1000
0 10 20 30 40Spacing between frames us
Rec
v W
ire r
ate
Mbi
t/s
50 bytes
100 bytes
200 bytes
400 bytes
600 bytes
800 bytes
1000 bytes
1200 bytes
1400 bytes
1472 bytes
gig7-jiveg1_UKL_31May05
0
20
40
60
80
100
0 10 20 30 40Spacing between frames us
% P
acke
t lo
ss
50 bytes
100 bytes 200 bytes
400 bytes 600 bytes
800 bytes 1000 bytes
1200 bytes 1400 bytes
1472 bytes
Connect. Communicate. CollaborateResults of Jodrell-JIVE eVLBI Transfer Tests
• Overall link quality Manchester/Jodrell to JIVE very good – near line rate possible
• Occasional negligible losses, very infrequent noticeable losses
• Packet re-ordering observed• Monitoring equipment will help to discover point in the link
where losses occur• These tests had no ‘exclusive’ access to machines –
possible for other users to have an effect• Real eVLBI sessions should be able to control access of
machines
Connect. Communicate. Collaborate
Conclusions and the FUTURE
• Data rates of 256Mbps feasible on real eVLBI runs• Data rates of 900Mbps possible with UDP protocol• The EXPRes EU eVLBI proposal: 16 telescopes in Europe,
Asia, Australia, Latin America and North America JBO-Onsala test link at 4 Gbps, submitted March 19th 2005
• eVLBI tests in EVN at ~ 1 per month• Network tests at up to 950 Mb/s – statistics of packet loss• Investigate alternatives to TCP, UDP – tsunami, VSI-E etc.• ESLEA tests on MB-NG and UKLight• eVLBI will be routine in 2006!!
Connect. Communicate. Collaborate
Backup Slides
Connect. Communicate. Collaborate
•Packet loss distribution:
( )t
p d
Cumulative distribution
Cumulative distribution of packet loss, each bin is 12 sec wide
Long range effects inthe data?
Poisson
Connect. Communicate. Collaborate
Divide time series of packets into 1000 slices of 50 packets
Total lost packets 1410 Average number / slice =
1.4
Calc Poisson Probability P(n, µ) = µ n e -µ
n!
Curves close but not exact Could be more than 1
process
VLBI Project: Packet Loss – Is it Poisson?
0
50
100
150
200
250
300
350
400
0 5 10 15n num lost in sub-sample
N(n
)
run12b
1
1.3
1.4
1.8
Connect. Communicate. Collaborate
Aggregated Variance Method Divide time series length N into
blocks of size m Calc mean of each section
Xm(k) k= 1 … N/m Calc variance VXm of these
Xm(k) Vary m size of the blocks
Plot on log-log & fit slope β Hurst parameter H
β = 2H -2 Measure:
β = -0.355 H 0.822 H =1 no long range dependence
VLBI Project: Packet Loss – Long Range Effects?
y = -0.355x + 2.8826
0
0.5
1
1.5
2
2.5
3
3.5
4
0 0.5 1 1.5 2 2.5 3sub-sample size Log10( m )
Agg
rigat
e-va
rianc
e L
og10
( X(m
) )