eVLBI at 1 Gbps unlimited networks

1
e-VLBI at 1 Gbps --unlimited networks?

• Tasso Tzioumis
• Australia Telescope National Facility (ATNF)
• 4 November 2008

2
Introduction

• Very Long Baseline Interferometry (VLBI)
• Combining (correlating) very distant radio
  telescopes
• virtual large telescope ? more resolution
  detail
• Baseline distance ? 100s-1000s km
• Interferometry pair-wise correlation
• Telescopes widely distributed over countries,
  continents, even in space.
• Traditionally record data on tapes or disks
  correlate later (days or weeks or months)
• e-VLBI
• Real-time VLBI using fast network transfers into
  hardware or software correlators

3
LBA Radio Telescopes in Australia
x Katherine
x ASKAP
x Yarragadee
x New Norcia
x
4
VLBI Astronomy Arrays (c. 2008)
o
o
o
o
o
o
o
o
o
o
o
x
o
o
o
o
o
O New telescopes in last decade
5
Resolution (Details)
Atmosphere gives 1" limit without corrections
which are easiest in radio
Jupiter and Io as seen from Earth 1 arcmin
1 arcsec 0.05 arcsec 0.001
arcsec
Simulated with Galileo photo
6
Cen A
7
Sensitivity

• Observe weak objects
• Look into different populations
• e.g. Hubble Deep Field
• View much older objects
• ? Look back in time
• New insights into universe
• Need similar sensitivity in the
• radio spectrum

HST Deep Field
8
Radio Telescope sensitivity

• Sensitivity depends on
• Size of the telescope - very expensive to
  increase
• Radio receivers limit of state of the art
  electronics
• Cryogenically cooled to liquid He temperatures
• Integration time limited by clock stability
• Bandwidth ? Data rate (after sampling)
• Most cost effective to achieve!?
• Most progress in recent years
• gt1 Gbps routinely achieved
• VLBI data on disks
• High data rates expensive to support

9
e-VLBI real-time observing over fast networks

• e-VLBI
• Transport the VLBI data over fast networks for
  real-time operation ? fast response
• Data rates at 1 Gbps or more are required
• BUT very expensive commercially
• ? Collaboration with network providers
• National Research and Education Networks (NREN)
• E.g. Internet2 DANTE collaboration AARNet
• Research in Astronomy Networking
• Very high and sustainable datarates (gt 12 hours)
• Testing the speed and reliability of the networks

10
e-VLBI Challenges - 1

• last-mile fibre connectivity
• Connect remote telescopes to the NREN backbone
• Fibre build needed - expensive
• But many institutions already at 1 Gbps
• Tools and protocols
• Sustained high rate traffic required
• TCP/IP needs fine tuning to achieve high rates
• Congestion control can create problems
• Other protocols needed (UDP modified TCP UDP)
• Research and development needed

11
e-VLBI Challenges - 2

• Long intercontinental distances
• RTT Australia-Europe gt 300 ms
• Routed Ethernet not consistently reliable
• Dedicated light-paths at 1 Gbps
• Can be built over ethernet backbone
• Require NREN cooperation over many countries
• e.g. 3 x 1 Gbps Oz-Holland in 2007
• Real-time data processing
• Hardware correlators new interfaces
• Software correlators new developments
• Supercomputer clusters and grid applications

12
e-VLBI Achievemnts - EXPReS project
connectivity world-wide
13
Oz achievements

• Telescope connectivity at 2 x 1 Gbps
• 2M fibre to telescopes built by CSIRO
• University fast connections in Uni Tasmania,
  Swinburne (Melbourne) Curtin (Perth)
• 1 Gbps operation demonstrated within Australia
• 512 Mbps production e-VLBI
• 512 Mbps operation with Europe (12 hours) in 2007
• 512 Mbps operation with Japan and China in 2008
• Internet2 IDEA award - 10 Gbps link in US for 1
  year
• New fibre-build for new telescope (ASKAP) in WA
• Construction to commence shortly
• Research for next generation radio astronomy
  instruments ( e.g. SKA 2B international
  project)

14
Image created by Paul Boven Satellite image Blue
Marble Next Generation, courtesy of NASA
Visibible Earth
15
Next steps

• 10 Gbps connectivity
• NREN backbones already at 10 Gbps
• Multiple colours possible (many ?? on one fibre)
• Dynamic circuit allocations
• Build light-paths and other circuits
  interactively
• Systems in test
• Distributed correlation
• grid-like applications
• Flexible operations on software

16
Lessons learned

• NREN collaboration critical
• Must involve networking community
• Symbiotic relationship science networks
• Network tools development
• Tools and protocols need development
• Sustained high data rates over long distances
  still problematic ? light-paths
• Very high data rates are achievable now
• Almost infinite data pipes
• NREN test circuits at low cost
• e.g. AARNet try-before-you-buy scheme
• 6-12 months free 1 Gbps connectivity!

17
Paradigm shift unlimited networks??

• Current way of science planning
• Current plans often limited by scarce network
  resources
• Researchers assume connectivity the limiting
  factor
• Often limit our own visions and horizons
• ? New way of planning?
• Plan on the science needs assuming almost
  infinite networks!!
• Fast research networks are arriving very quickly
• Expand our horizons now!

18
Thank you
ATNF Tasso Tzioumis LBA eVLBI Phone 61 2
9372 4350 Email Tasso.Tzioumis_at_csiro.au Web
www.atnf.csiro.au/vlbi