Presentation Title Size 30PT

Comments

Transcription

Presentation Title Size 30PT
Ultra High-speed
Telecommunication
Requirements for the SKA
Peter Elford
Business Development manager, Higher Education and Research
[email protected]
© 2011 Cisco and/or its affiliates. All rights reserved.
1
• Where are the traffic flows and their sources and sinks ?
• What are the size of the traffic flows ?
• Over what distance do the traffic flows travel ?
• What are the interfaces into/out of sources and sinks ?
• Will the requirements change over time ?
Drives flexibility and growth design considerations
• What tradeoffs are possible between network/compute/storage ?
• What is the cost of bandwidth ?
Do you own the network infrastructure or are you buying services ?
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
2
• SKA is still largely a “science project”
Networking requirements are still TBD/TBC ... many estimates
Lots of focus on SKA2 ... vs. SKA1, and a ~50 year operational life
• The next 2-3 years will transform SKA into an engineering project
Challenge will be balancing science vs. engineering
• Role of industry seems clear
“Many of the SKA’s characteristics can only be delivered if new technologies
are developed specifically for the SKA, and use is made of technologies
constantly being renewed and developed in industry such as high
performance computing and high speed telecommunications transmission
systems”
“Industry will play a crucial role in the delivery and through‐life support
of the SKA technologies and infrastructure. The scale of the SKA, and the
need to produce components, requires industry engagement on a scale
unprecedented in radio astronomy. The involvement of organisations with
experience and expertise in delivering demanding technological
specifications within a production cost envelope will be essential.”
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
3
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
4
• Multiple network needs
Sensor
(00’s – 000’s)
Timing
Custom Built
Computer
(1)
Remote, RFI Quiet
Data Centre
(1)
Real time,
Uni-directional
Data Flow
Signal
Processing
Facility
Science
Computation
Facility
Integration ...
... at least at layer 1 ?
• Obvious North-South
and East-West flows
Analysis,• The big network is BIG ...
Science
Science
NREN
(~150)
Public
Internet
(1)
Control
Operations
SCADA
Power
Major City
© 2010 Cisco and/or its affiliates. All rights reserved.
Data
Objects
World
Cisco Confidential
5
Bandwidth
Remote, RFI Quiet
Signal
Processing
Facility
Distance
Mostly Campus/Metro
(5-180 km)
and
Continental
Long Haul for SKA2
(up to 3,000 km)
Long Haul
(100’s of km)
Science
Computation
Facility
© 2010 Cisco and/or its affiliates. All rights reserved.
Major City
Data Centre
(10’s of m)
Cisco Confidential
6
• Multiple dish types
Dish (High)
Physically very different
Each type captures collects
different frequencies
• Dishes can have two sensors
Single Pixel Feed (SPF)
Phased Array Feed (PAF)
Sparse Aperture Array (Low)
© 2010 Cisco and/or its affiliates. All rights reserved.
Dense Aperture Array (Med)
Cisco Confidential
7
SKA Phase 2
Possible Configuration
SKA Phase 1
Possible Configuration
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
8
Dishes:
Core: 20%
Inner: 30%
Aperture Arrays:
Core: 30%
Inner: 36%
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
9
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
10
.
.
Boolardy
Station, MRO
© 2010 Cisco and/or its affiliates. All rights reserved.
NewNew
Zealand
Zealand
Cisco Confidential
11
• Sensor to Signal Processing Facility
SKA1
Sensor
Bit Rate
(Gbps)
SKA2
n
Total
Bandwidth
(Tbps)
Bit Rate
(Gbps)
n
Total
Bandwidth
(Tpbs)
3,000
648
Dish with SPF
24
250
6
216
Dish with PAF
929
50
45
929 ~2,500
AA-low
1,216
250
304
33,440
250
8,360
AA-mid
N/A
0
0
16,800
250
4,200
TOTAL
356
2,322
15,530
ASKAP
Sensor
Bit Rate
(Gbps)
n
Total
Bandwidth
(Tbps)
1920
36
69
Dish with PAF
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
12
• Signal processing needs to be located “close” to centre of dish/sensor
array, but needs to be 10’s km away to avoid RFI (maybe 10-20km)
• Huge number of sensor connections means cost needs to be very low
Fast less important than cheap ... Many 10G vs. fewer 100G (say)
Design needs to balance cost of fibre vs. deployment cost vs. cost of optics
Fibre density will be challenging particularly for SKA2
• Signal processor and sensors are both “custom-built”
Granularity of data streams needs to be standardised (1G, 10G, 100G, ???)
Per sensor transmission will be a bundle of data streams
• Most sensors are within “campus” short-haul distance of signal processing
More distant sensors may need amplification
• Remote stations will have to be DWDM long hauled to signal processor
Few remote stations means this will be a relatively
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
13
• Signal Processing Facility to Science Computing Facility
SKA1
3 Tbps
SKA2
138 Tbps (?)
ASKAP 0.16 Tbps
• Very, very significant reduction in bandwidth
• Well within capabilities of commercial offerings in time required
Many long haul DWDM systems with 80 channels of 100G = 8T
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
14
Data
Centre
Residential
Broadband
(FTTH)
Internet
Cloud
(Industrial
Computing)
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
15
DELL'ORO
GROUP
$1,000
GE ASP
10GE ASP
Dell’Oro Ethernet
Forecast 2011-2015
Vol. 17. No. 1
*Average
Sell
Price
$0
2003
© 2010 Cisco and/or its affiliates. All rights reserved.
2005
2007
2009
2011
2013
2015
Cisco Confidential
16
• Technology exists to meet requirement – just requires cost optimisation
Driven by factors other than SKA
• Interface standards critical to support loose coupling and, through
alignment with industry standards, cost minimisation
• Many forces driving telecommunications industry towards SKA-relevant
capabilities and price points
• Unless a clever incremental approach can be identified, may have to
build SKA2-ready passive infrastructure up front
• Very different networking architectures need to be applied to different
elements of the SKA solution
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
17
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
18
© 2010 Cisco and/or its affiliates. All rights reserved.
Cisco Confidential
19

Similar documents