MASSACHUSETTS INSTITUTE OF TECHNOLOGY
HAYSTACK OBSERVATORY
WESTFORD, MASSACHUSETTS 01886
Telephone:
Fax:
978-692-4764
781-981-0590
8 July 2003
TO:
Distribution
FROM:
Alan R. Whitney
SUBJECT:
7 July 2003 e-VLBI telecon summary
Attendees:
Bill Fink, Lee Foster, Pat Gary, Chuck Kodak, Kevin Kranacs– GSFC
Tom Lehman – ISI-E
Jerry Sobieski - MAX
Rick Larkin, Peter Schulz – MIT Lincoln Lab
Charles Yun – Internet2
Kevin Dudevoir, David Lapsley, Arthur Niell, Mike Titus, Alan Whitney – Haystack Observatory
This telecon is one of an ongoing series of telecons to prepare for gigabit/sec e-VLBI demonstrations
between NASA GSFC and MIT Haystack Observatory using a combination of network facilities
including all or part of Glownet, Bossnet, ISI-E, SuperNet, Max and GSFC/HECN.
ACTION ITEMS ARE HIGHLIGHTED IN RED.
Glownet/Bossnet
Rick re-iterated that the fuding future of Bossnet look very promising for the next four years.
Peter reported that the OC-48 status is still the same as last report. Peter and Tom plan to do tests
soon.
GigE has been stable on Bossnet.
NASA/GSFC
The link between GSFC and MAX has been upgraded to include an aggregated 2xGigE link (links
H11 and K3).
Bill reported four new 10GigE NIC’s have been received; testing hopefully to begin soon.
GGAO has received Mark 5A and is ready to resume testing.
GigE switch testing
Switch testing is continuing, but neither Paul Lang (who has been testing the SMC 8624T switch) nor
Russ Roberge (who has been testing the Dell 5224) were available to give us a complete update. Bill
Fink reported that he understood that testing of the SMC switch was successful in pushing full datarate simultaneously through all 24 ports. Rick indicated that the problems with the Dell switch
reported by Russ in an earlier telecon are still outstanding as far as he knows.
minutes29.doc 7/9/2003
1
Japan, Hawaii and German Connections
David reported on a successful e-VLBI experiment using Kashima and Westford antennas on 28
June. After a 2-hour experiment, ~80 GB of data were transferred, correlated, and then reduced for
earth-rotation results within less than 24 hours; there were a number of little hitches, and we think we
can easily reduce the turnaround time to <8 hours in the future. The trans-Pacific link was SuperSinet, which was made available especially for this experiment; average transfer speeds were about
47 Mbps Haystack-to-Kashima and 107 Mbps Kashima-to-Haystack, which is much better than we
have been getting in the past, but still nowhere near what we should be getting (Super-Sinet is OC-48,
with a known bottleneck of OC-12). UDP transfer rates were measured with iperf to be 200-300
Mbps with ~1% packet loss. Post-experiment test with ‘high-speed’ TCP/IP (‘HSTCP’, from Sally
Floyd at ICIR; see http://www.icir.org/floyd/papers.html) showed essentially no speed gain,
apparently because the packet loss rate was too high to allow the high-speed algorithms in HSTCP to
kick in.
The trans-Pacific connection has been re-established through GEMnet, but plans are underway to
connect via TransPAC. David is working with Chris Robb at Indiana U. and Hirabaru-san of CRL
(who has a server on the Japanese end of TransPAC) to do testing with Haystack, Abilene and Japan.
Hirbaru-san is working toward getting Kashima connected to TransPAC at high speed.
Kevin D. reported that the problems in the Hawaii connection are finally being looked at by the
PMRF people; the problem is believed to be in the 3xDS3 microwave link from PMRF to Kokee
controlled by PMRF. No significant change in observed rates: ~5Mbps TCP and ~20Mbps UDP in
both directions, but with always a rate-independent ~25% packet loss in Kokee-to-Haystack
direction. Hopefully there will be good news and improvements within the next week. Kevin will
work with Tom to create a detailed path diagram to Hawaii.
There has been little progress on the German connection, though Wettzel is seeking funding of a
direct 34 Mbps link to Wettzell which, if funding is secured, could be in place within a few months.
Server at U. of Regensburg should become available again this week to resume some testing.
Performance Monitoring
Andy Germain has been with Pat, Tom and Kevin with regard to installing his performancemonitoring software on various test workstations, including ‘kame’ at ISI-E, ‘superglide’ at LL and
‘evlbihay’ at Haystack. Scheduling of Bossnet may preclude Haystack from regular testing for the
near term, but testing should begin on others where possible.
Charles reported his plan for a deployment of performance-monitoring software and hardware, which
has many parallels. These will include WEB100 testing, including distribution of some I2
‘cakeboxes’ which are test workstations specially configured for performance testing and monitoring
(though not capable of Gigabit/sec speeds). Charles has been coordinating with Steve Parsley at
JIVE, ?? in Chile and David at Haystack. Because Haystack is dependent on scheduling of Bossnet,
making it difficult to do on a regular basis, Charles suggested that perhaps testing could perhaps be
moved further upstream to some place like ISI-E; Charles and Tom will consult about the
possibilities of installing a cakebox at ISI-E. Charles said I2 will process and post the data for web
access, and use the data for determining what further testing strategies should be pursued.
Charles is seeking contacts in Japan to allow similar testing to Japan.
minutes29.doc 7/9/2003
2
Some coordination of these two testing programs will need to take place so that they don’t collide and
produce spurious results. Tom and Charles will coordinate; Pat Gary will forward Andy Germain’s
to Charles so Andy can be included in the coordination efforts.
Newest version of iperf allows for 2-way tests, so all tests may be initiated from the server side alone.
Bill Fink has suggested a network delay simulator with controllable loss rates would be a good tool
for performance testing; he is working on one. Kevin Dudevoir and David Lapsley are also working
on a similar simulator.
Next telecon
Next telecon is scheduled for Mon, 28 July 2003 at 3 pm EDT (note: 3 pm, not 2 pm)
xc:
Steve Bernstein, LL
Jim Calvin, LL
Rick Larkin, LL
Lorraine Prior, LL
Peter Schulz, LL
Leslie Weiner, LL
Herbert Durbeck, GSFC
Bill Fink, GSFC
Lee Foster, GSFC
Pat Gary, GSFC
Chuck Kodak, GSFC
Kevin Kranacs, GSFC
Paul Lang, GSFC
Aruna Muppalla, GSFC
Bill Wildes, GSFC
Dan Magorian, UMCP
Tom Lehman, ISI-E
Jerry Sobieski, MAX
Guy Almes, Internet2
Charles Yun, Internet2
Richard Crowley, Haystack
Kevin Dudevoir, Haystack
Hans Hinteregger, Haystack
David Lapsley, Haystack
Arthur Niell, Haystack
Joe Salah, Haystack
minutes29.doc 7/9/2003
3
Req’d during data acq
A0
Mk5 Cntl
Cmptr
A1
Mark 5
Westford Antenna
Millstone RPE Bldg
A2
SX
1
Summit
1i
GigE
SX
Westford
LAN
Mark 5
Dedicated no scheduling
A2A
Alpine
A4
3804
Haystack Correlator
Internet
B1
Mark 5
(correlator)
Haystack
LAN
B1A
Corr
Cntl
Cmptr
SX
28-1A
B2
GigE
B5
14
SX 28-1A
RJ-45
13
2
B4A
1
LL
C2
Dedicated no scheduling
Mark 5 B2A
(correlator)
Mark 5
B3A (correlator)
C3
16
Summit
WDM
GigE
5i
15
(Shared with LL on
scheduled basis)
Cisco
1500
B6
RJ-45
B11
B3
Test
WS
Req’d during data processing
B4
Congress St
Boston
(~50 km)
Amp
D3
WDM
D2
Lincoln Lab
Schedule to be
coordinated with LL
Cisco
1500
3l (2 GigE + 1 OC-48)
Dell Poweredge 2500
(’evlibhay’)
GlowNet
GlowNet
D1
Bossnet
Scheduling on
Bossnet calendar
Shared with MIT traffic
E6
(~100 km)
GigE
E3
E2
GigE
E1
E5A
GigE
8
A3
A1A
Schedule to be
coordinated with LL
C1
E4A
Test
WS
Juniper
M20
Alpine
3804
E7
E4
GigE
Test
WS
GigE
E5
MIT
Campus
OC-48
WDM
1310nm
F1
E8
Amp
Amp
F2
E9
Other users (no control)
To be upgraded to OC-48
(’superglide’)
(’evlbill’)
From ISI-E
Dell Poweredge 500SC
(provided by Haystack)
Figure 1: e-VLBI Path - Haystack to ISI-E
evlbi011.cdr
27 Mar 03
ISI-E
Test
WS
Scheduling on
Bossnet calendar
MAX@ISI-E
G6
G7
Test
WS
Summit
5i
GE
(’kame’)
G13
GigE
G1
G8
G2
OC-48
WDM
G6B
G7B
G5
(’omega’)
Bossnet
GE
G6C
GigE
GigE
OC-48
G16
Abilene
G14
MAX@Eckington
H7
Abilene
Juniper
M160
OC-48
H7B
GigE
H8
GigE
H9
Primary e-VLBI path
H7C
DREN
LX (~14km)
Need details
H2
GigE
G11
H4
H3A
J2
H3
No scheduling;
sufficient capacity
for e-VLBI
GigE
Juniper
M160
H6
OC-48
GigE
H6B
J3
Test
WS?
OC-48
Borrowed OC-48
interface - may not
be able to keep
G10
MAX@UMCP
H7A
H1
G12
GigE
Test
WS?
H5
Frame relay encapsulation
No jumbo frames
H7C
Juniper
OC-48
M160
Switch
Backup plan
Juniper
M40
G15
GigE
Nortel
8600
Summit
5i
G4
MAX@GWU
to be upgraded to Cisco 6509
Cisco
4912
Switch
J4
GigE
J5
WDM
GigE
on loan from GSFC
Summit
5i
H10
2xGigE H11
LuxN
AMP
LuxN
H6A
H6C
J1
USNO?
J6
GSFC/Bldg 28
K4
K1
WDM
G4 Mac with
Yellowdog Linux
(’clifford’)
GGAO/Bldg 201
GigE
Summit
K2
5i
2xGigE
K3
K5A
K5
Test
WS
Test
WS
K6
K6A
GigE LX (~7.25km)
1310nm K7
K5B
Cisco
12016
K6B
GGAO Antenna Trailer
GigE Summit L3
GigE
L1
5i
L5
G4 Mac
(’pluto’)
Windows PC
(’xly’)
RJ-45
Test
WS
Internet
L4
L2
Dedicated - no
scheduling necessary
(~500m)
RJ-45
Mark 5
NASA
LAN
L3A
RJ-45
L4A
Mark 5
L6
Mark 5
Cntl
Cmptr
L5
Req’d during data acq
evlbi012.cdr
Figure 2: e-VLBI Path - ISI-E to GSFC/GGAO
8 July 03