NREN 3.0STEPHEN WOLFF
Internet2
JUNE 2013
•  Mobile Internet
•  Automation of knowledge work
•  Internet of things
•  Cloud technology
•  Advanced robotics
•  Autonomous/near-autonomous
vehicles
•  Next-generation genomics
•  Energy storage
•  3-D printing
•  Advanced materials
•  Advanced oil/gas exploration,
recovery
•  Renewable energy
[ 2 ]
Twelve disruptive technologies
June 11, 2013 © 2013 Internet2
“Disruptive Technologies: Advances that will transform life, business, and the global economy”
McKinsey Global Institute, 2013
These technologies are forecast to account for $33 trillion of the global economy in 2025
[ 3 ]
Big Data is the primary driver
June 11, 2013 © 2013 Internet2
Familiar numbers:
•  Major genome sequencing
centers @ 2 PB/year
•  LSST @ 10 PB/year
•  LHC @ 30 PB/year
•  SKA @ 1 EB/day
•  …
•  Continued rise of “team” science
“To a very considerable extent statistics and applied mathematics and parts of
pure mathematics have become ‘group’ science requiring bodies of graduate
students, postdocs and computing resources…”
http://www.nsf.gov/attachments/124926/public/DMS_Name_Change_Committee_Report_Final_4-1-12.pdf
•  One Culture. Computationally Intensive Research in the Humanities and
Social Sciences, Christa Williford & Charles Henry, CLIR, June 2012
•  Internet2/Microsoft workshop on CI for social science February 2013
•  IEEE Workshop on Big Humanities October 2013
[ 4 ]
Then there’s the long tail
June 11, 2013 © 2013 Internet2
Globalization of scientific research, e.g.:
•  LHC involved scientists from 36 countries
•  Annual funding for India-US collaborative research exceeds
$2B, and there are more than 360 projects funded at
$250,000 or more
•  Annual funding for China-US collaborative research exceeds
$1B, and there are more than 450 projects funded at
$100,000 or more
[ 5 ]
…and ‘teams’ are not all local
June 11, 2013 © 2013 Internet2
Globalization and democratization of higher education –
•  Remote campuses
•  Classical distance education
•  MOOCs
[ 6 ]
Nor is higher education…
June 11, 2013 © 2013 Internet2
[ 7 ]
US NREN speed vs. Moore’s Law
June 11, 2013 © 2013 Internet2
1985 1990 1995 2000 2005 2010
YEAR
10 kb/s
100
1 mb/s
10
100
1 Gb/s
10
100 Gb/s
[ 8 ]
OAM multiplexing
June 11, 2013 © 2013 Internet2
[ 9 ]
OAM multiplexing
June 11, 2013 © 2013 Internet2
•  Willner et al USC
•  Nearly 100 b/s/Hz
BUT
•  You can’t beat
Shannon
Moving “Big Data” has led to fundamental advances in the
understanding and design of networks
•  Van Jacobson and Berkeley UNIX source (1988)
•  Matt Mathis and Web100 (2003)
•  Science DMZ (ESnet, 2009)
[ 10 ]
Big Data
June 11, 2013 © 2013 Internet2
Using “Big Data” has led to fundamental advances in
science and in analytics
•  “Traditional” science, with Big Data (e.g., LHC)
•  The Fourth Paradigm – learning algorithms, coping with
velocity, unstructured data, visualization,…
[ 11 ]
Big Data
June 11, 2013 © 2013 Internet2
•  Globalization ⇒ No NREN is an island
•  Team science ⇒ (possibly virtual) team networks for
virtual communities
•  Big Data ⇒ More than just bigger pipes
[ 12 ]
Implications/consequences
•  Globalization ⇒ It’s time to engineer global R&E
connectivity: from NRENs to a GREN
•  Team science ⇒ A one-size-fits-all IP network is not
enough. SDN (e.g., OpenFlow, OSCARS) will help users
and applications configure the network
•  Big Data ⇒ Re-architecting (Science DMZ), tuning
(Web10G), measuring (perfSONAR)
[ 13 ]
Further implications
•  Data = “content”
•  Future NREN node?
•  CDN/NDN
–  via SDN?
[ 14 ]
…and a speculation
Source: D. Comer
•  Growing number of instances where the data are too
numerous to move
•  Instead, move the computation to the data
•  Computation in place
–  AWS, Microsoft Azure, cloud computing generally…
•  Map-reduce and Hadoop
–  Widely applicable and successful programming model
–  Not well adapted to typical HPC architectures
[ 15 ]
The Black Swan
•  All the data are not – and cannot – be in one place
–  10 EB of spinning disk ≈ 2.4 Gwatts
not counting backups or redundancy measures"
•  “Computation in place” => GREN as backplane
[ 16 ]
Distributed data
•  Computers, datastores, and networks are all programmable
•  Computational models / multicomputer architectures, network
architectures, and storage architectures are converging
disciplines
therefore
•  NREN engineers and operators must acquire added skills.
•  The need is for support staff who are versed in computational
_______, and able to program everything (see above)
[ 17 ]
Convergence
•  “Big data” and the growth of team science are bringing in to our
community researchers with little background in computing
beyond their laptops.
•  We are faced with a group of scientists who are “…analyzing data
they did not collect, using software they did not write, on
computers they do not own.” – Myron Livny
•  Not only are Renaissance thinkers needed for support staff, we
must begin to hold scientific software to the same standards as
publications. cf. Joppa et al, “Troubling Trends in Scientific
Software Use”, Science, v 340, pp 814-815, 2013/05/17
[ 18 ]
Bootnote
Bootnote
NREN 3.0STEPHEN WOLFF
[ 19 ]
swolff@internet2.edu
June 11, 2013 © 2013 Internet2
Thank	
  you!	
  

NREN 3.0

  • 1.
  • 2.
    •  Mobile Internet • Automation of knowledge work •  Internet of things •  Cloud technology •  Advanced robotics •  Autonomous/near-autonomous vehicles •  Next-generation genomics •  Energy storage •  3-D printing •  Advanced materials •  Advanced oil/gas exploration, recovery •  Renewable energy [ 2 ] Twelve disruptive technologies June 11, 2013 © 2013 Internet2 “Disruptive Technologies: Advances that will transform life, business, and the global economy” McKinsey Global Institute, 2013 These technologies are forecast to account for $33 trillion of the global economy in 2025
  • 3.
    [ 3 ] BigData is the primary driver June 11, 2013 © 2013 Internet2 Familiar numbers: •  Major genome sequencing centers @ 2 PB/year •  LSST @ 10 PB/year •  LHC @ 30 PB/year •  SKA @ 1 EB/day •  …
  • 4.
    •  Continued riseof “team” science “To a very considerable extent statistics and applied mathematics and parts of pure mathematics have become ‘group’ science requiring bodies of graduate students, postdocs and computing resources…” http://www.nsf.gov/attachments/124926/public/DMS_Name_Change_Committee_Report_Final_4-1-12.pdf •  One Culture. Computationally Intensive Research in the Humanities and Social Sciences, Christa Williford & Charles Henry, CLIR, June 2012 •  Internet2/Microsoft workshop on CI for social science February 2013 •  IEEE Workshop on Big Humanities October 2013 [ 4 ] Then there’s the long tail June 11, 2013 © 2013 Internet2
  • 5.
    Globalization of scientificresearch, e.g.: •  LHC involved scientists from 36 countries •  Annual funding for India-US collaborative research exceeds $2B, and there are more than 360 projects funded at $250,000 or more •  Annual funding for China-US collaborative research exceeds $1B, and there are more than 450 projects funded at $100,000 or more [ 5 ] …and ‘teams’ are not all local June 11, 2013 © 2013 Internet2
  • 6.
    Globalization and democratizationof higher education – •  Remote campuses •  Classical distance education •  MOOCs [ 6 ] Nor is higher education… June 11, 2013 © 2013 Internet2
  • 7.
    [ 7 ] USNREN speed vs. Moore’s Law June 11, 2013 © 2013 Internet2 1985 1990 1995 2000 2005 2010 YEAR 10 kb/s 100 1 mb/s 10 100 1 Gb/s 10 100 Gb/s
  • 8.
    [ 8 ] OAMmultiplexing June 11, 2013 © 2013 Internet2
  • 9.
    [ 9 ] OAMmultiplexing June 11, 2013 © 2013 Internet2 •  Willner et al USC •  Nearly 100 b/s/Hz BUT •  You can’t beat Shannon
  • 10.
    Moving “Big Data”has led to fundamental advances in the understanding and design of networks •  Van Jacobson and Berkeley UNIX source (1988) •  Matt Mathis and Web100 (2003) •  Science DMZ (ESnet, 2009) [ 10 ] Big Data June 11, 2013 © 2013 Internet2
  • 11.
    Using “Big Data”has led to fundamental advances in science and in analytics •  “Traditional” science, with Big Data (e.g., LHC) •  The Fourth Paradigm – learning algorithms, coping with velocity, unstructured data, visualization,… [ 11 ] Big Data June 11, 2013 © 2013 Internet2
  • 12.
    •  Globalization ⇒No NREN is an island •  Team science ⇒ (possibly virtual) team networks for virtual communities •  Big Data ⇒ More than just bigger pipes [ 12 ] Implications/consequences
  • 13.
    •  Globalization ⇒It’s time to engineer global R&E connectivity: from NRENs to a GREN •  Team science ⇒ A one-size-fits-all IP network is not enough. SDN (e.g., OpenFlow, OSCARS) will help users and applications configure the network •  Big Data ⇒ Re-architecting (Science DMZ), tuning (Web10G), measuring (perfSONAR) [ 13 ] Further implications
  • 14.
    •  Data =“content” •  Future NREN node? •  CDN/NDN –  via SDN? [ 14 ] …and a speculation Source: D. Comer
  • 15.
    •  Growing numberof instances where the data are too numerous to move •  Instead, move the computation to the data •  Computation in place –  AWS, Microsoft Azure, cloud computing generally… •  Map-reduce and Hadoop –  Widely applicable and successful programming model –  Not well adapted to typical HPC architectures [ 15 ] The Black Swan
  • 16.
    •  All thedata are not – and cannot – be in one place –  10 EB of spinning disk ≈ 2.4 Gwatts not counting backups or redundancy measures" •  “Computation in place” => GREN as backplane [ 16 ] Distributed data
  • 17.
    •  Computers, datastores,and networks are all programmable •  Computational models / multicomputer architectures, network architectures, and storage architectures are converging disciplines therefore •  NREN engineers and operators must acquire added skills. •  The need is for support staff who are versed in computational _______, and able to program everything (see above) [ 17 ] Convergence
  • 18.
    •  “Big data”and the growth of team science are bringing in to our community researchers with little background in computing beyond their laptops. •  We are faced with a group of scientists who are “…analyzing data they did not collect, using software they did not write, on computers they do not own.” – Myron Livny •  Not only are Renaissance thinkers needed for support staff, we must begin to hold scientific software to the same standards as publications. cf. Joppa et al, “Troubling Trends in Scientific Software Use”, Science, v 340, pp 814-815, 2013/05/17 [ 18 ] Bootnote Bootnote
  • 19.
    NREN 3.0STEPHEN WOLFF [19 ] swolff@internet2.edu June 11, 2013 © 2013 Internet2 Thank  you!