Transcript NASA Tests e-VLBI concept from the Madrid Deep Space Network to

NASA
Tests e-VLBI concept from the Madrid Deep
Space Network to JPL
Thom Stone
Principal Computer Scientist
CSC-NASA Ames Research Center
Deep Space Network and
eVLBI
• What is “electronically linked-Very Long
Baseline Interferometry” and why do we need
it?
• What is the DSN and how is it connected
now?
• Our testing, why, when how
• Results
• Conclusions and suggestions
• Data from deep space - The future
What is e-VLBI
• eVLBI is the process of using high speed
networks to connect radio telescopes
separated by large distances (100-1000s of
km) instead of the traditional method of
recording onto magnetic tape and shipping
the recorded data to a central correlator
• In other words several radio telescopes
dispersed over the planet are tied together
over high speed networks to form one big
telescope.
Arecibo radio telescope
(Puerto Rico)
• World’s largest singledish radio telescope
• 305m diameter, 40K
perforated aluminum
panels
e-VLBI
• Tried over European networks successfully
2007
• Requires the high available bandwidth of a
advanced commercial or research network
• Main application is tracking spacecraft and
other objects in deep space (triangulation)
• Requires exacting time synchronization
between telescopes (1013 sec.)
• Logical method for creating large array
What is the NASA Deep
Space Network (DSN)?
• Three arrays of radio telescopes (dish
antenna) spread around the world to give
maximum sky coverage
• Used to contact spacecraft beyond earth orbit
both NASA and International
• TDRSS satellites are used to communicate to
spacecraft in Earth orbit
Deep Space Network Ground Stations:
•Near Barstow California in the Mojave desert
•Outside of Madrid Spain
•In Parks Australia
Why the DSN
• Each station has many antenna
including one 70 meter to:
– Send Beatles songs to other star system
– Tracking, commanding, and receiving data from
spacecraft (NASAs and International partners)
outside of Earth orbit. Some are light hours away
– Some radio Astronomy
Current DSN connectivity
• Minimal Internet
connectivity at each site
even though they are near
major high speed research
networks(as well as
commercial networks)
• Use leased lines and tapes
to move data to NASA
centers and then on to
researchers
NREN-Madrid-JPL Test
• Took place in December 2006 after several months
of discussion and design
• Goal was to form eVLBI between Madrid and
Goldstone DSN stations with a Beowolf cluster (JVC)
at JPL as correlator
• Object was to demonstrate that data now sent on
tapes to JPL could utilize HPREN (High Performance
Research and Education Networks) instead (faster,
and better)
• Cooperation between JPL, ARC, REDIRIS, and the
DSN
• Up to a terabyte a day can be generated in Madrid
JPL Correlator
(Beowulf Cluster)
Limiting Factors
• There is only an E1 link (2 Mb/sec) between
Madrid DSN and REDIRIS the Spanish
HPREN in downtown Madrid
• The default route to JPL from RedIRIS is via
commercial network not NREN’s 1 Gbps link
• The software/hardware (MARK-5 system)
used to make tapes was not set up for high
speed transmission of the data
Test Set Up
•
•
•
•
A “Mark-5” computer, used to generate files from data received from
the antenna from spacecraft was moved to the REDIRIS operations
center on a Gigabit LAN segment
Success criteria was to prove that 1/2 Tbyte could be moved in 18
hours or less (about 60-100 mbits per second with TCP average)
A Route was set up between the Mark-5 computer at REDIRIS to the
Columbia Supercomputer
UDP and TCP tests were performed between hosts
– Between hosts on the same LAN segment
– Between hosts on the wide area
– Using the Mark-5 software
Route
• Mark-5 connected to RedIRIS LAN (1
Gbps)
• RedIRIS connects to GEANT
(European backbone) 1 Gbps
• GEANT Spain to GEANT France to
transatlantic cable (10 Gbps)
• GEANT to Abilene (10 Gbps)
• Abilene to NREN (10 Gbps)
• NREN would need to set up route
from RedIRIS to machine at JPL for
real test
• 40km Madrid to RedIRIS: work in
progress
Test Results
• Used IPERF etc. for testing
• Able to get up to 800 Mbps UDP with a
few errors from Spain (RedIRIS) to
ARC. Less loss ARC->Spain
• Got 60-90 Mb/sec TCP depending on
time of day without optimization
• Unable to move files using MARK-5
software
Conclusion and Suggestions
• This is a viable concept, data can be moved
faster and cheaper
• Improvement can be made with Jumbo
Frames and multi-stream or other fast TCP
• Host tuning is a must
• A great candidate application for scavenger
(less than best effort) service
What’s Up
• European VLBI up and working
• 40 KM Link from Madrid DSN to RedIRIS not
funded even though pricing cheaper than low
speed point to point to JSC/JPL
• Reason is funding and distrust of IP networks
by some in NASA
• Whole DSN in need of upgrade to support
future space exploration
What’s Next
• Keep requesting funding for high speed
links to HPRN from all three DSN sites
• New NASA lunar endeavors will require
higher data rates and faster access to
data sooner rather than later
Un-Indicted Coconspirators
• "Garcia Miro,
Cristina"
•
<[email protected]
asa.gov>
Un-Indicted Coconspirators
•
•
•
•
•
•
•
•
" Fernando Aragon" [email protected]
"Cristina Garcia Miro" [email protected]
"Raul Alonso" <[email protected]>
"Charles J. Naudet” [email protected]
“Tom Kuiper” [email protected]
[email protected]
Ken Freeman, Ray Gilstrap, Hugh Lamaster at ARC
Others at Madrid DSN and JPL
Thank you
Questions? Comments?
Thom Stone
NASA Ames Research Center / CSC
thom.stone @ nasa.gov