Transcript
Clouds and Optical
p
Networks
Bill St. Arnaud
CANARIE Inc – www.canarie.ca
[email protected]
Unless otherwise noted all material in this slide deck may be reproduced,
modified or distributed without prior permission of the author
Taxonomy of Clouds
> Application or platform clouds
– Google
g Apps,
pp SalesForce
– No specific allocation of resources or location
> Computational Clouds
– Amazon EC2
– Specific allocation of storage and computation – no specific
location
> Private Clouds
– VmWare Datagardens,
Datagardens Grids
– Specific allocation of storage and computation at a specific
number of locations
Bandwidth demand for
clouds
> In January of 2008 Amazon Web Services consumed more
bandwidth than do the entire global network of
Amazon.com retail sites.
Clouds and Climate Change
> Clouds will be essential as we move to a zero carbon
Internet
> Computers will become handheld and solar powered
> Applications and services will be hosted on the
cloud that is based on zero carbon infrastructure
> Businesses will move services and applications to
zero carbon data centers using clouds
> Optical networks will be critical to sustain a zero
carbon cloud
Many examples
Wind powered data centers
Ecotricity in UK builds windmills at
data center locations with no capital
cost to user
Hydro-electric powered data centers
Relocation of Nordic HPC facilities to
Iceland
Data Islandia
Digital Data Archive
ASIO solar powered data centers
17
Zero Carbon strategy
essential
> Zero carbon strategy using renewable energy critically important if
governments mandate carbon neutrality, or if there is a climate
catastrophe
> With a zero carbon strategy growth in demand for ICT services will not
effect GHG emissions
– Anything times zero is always zero
> Wind and solar power are most likely candidates because of
opportunity cost/benefit analysis especially time to deploy
– Nuclear has high opportunity cost because of time to deploy
– http://climateprogress.org/2008/12/14/stanford-study-part-1-wind-solarbaseload-easily-beat-nuclear-and-they-all-best-clean-coal/
> But renewable energy sites are usually located far from cities and
electrical distribution systems are not designed to carry load
– http://www.americanprogress.org/issues/2008/12/pdf/renewable_transm
http://www.americanprogress.org/issues/2008/12/pdf/renewable transm
ission.pdf
15
“Zero
Zero Carbon”
Carbon ICT
> Purchasing green power locally is expensive with significant transmission
line losses
–Demand for green power within cities expected to grow dramatically
> ICT facilities DON’T NEED TO BE LOCATED IN CITIES
–-Cooling also a major problem in cities
> But most renewable energy sites are very remote and impractical to
connect to electrical grid.
– Can be easily reached by an optical network
– Provide
P id iindependence
d
d
ffrom electrical
l t i l utility
tilit and
d hi
high
h costs
t iin wheeling
h li power
– Savings in transmission line losses (up to 15%) alone, plus carbon offsets
can pay for moving ICT facilities to renewable energy site
> ICT is only industry ideally suited to relocate to renewable energy sites
– Also ideal for business continuity in event of climate catastrophe
16
Optical networks, SOA and
Virtualization are key
> Optical networks (as opposed to electronic routed
networks) have much smaller carbon footprint
> Significant reduced CO2 impacts are possible through
use of SOA, clouds, web services, virtualization,
UCLP, grids, Web 2.0, NGI etc.
> Computer and networks architectures to connect
remote computers
computers, databases and instruments will be
essential for zero carbon ICT
CANARIE Green-IT Pilot
> $3
$3m - $4m
$4 allocation
ll
ti for
f Green
G
cyber-infrastructure-IT
b i f
t
t
IT pilot
il t
testbed
> Two objectives:
– Technical viability and usability for relocating computers to zero
carbon data centers and follow the sun/follow the wind grid
– Business case viability of offering carbon offsets (and or
equivalent in services) to IT departments and university
researchers who reduce their carbon footprint by relocating
computers and instrumentation to zero carbon data centers
> International partnership with possible zero carbon nodes
using virtual router/computers in Spain, Ireland, California,
A
Australia,
li British
B i i h Columbia,
C l bi Ottawa,
O
Quebec
Q b and
d Nova
N
Scotia
S i
25
PROMPT – Next Generation Internet to
Reduce Global Warming
Research on router,
router optical,
optical W/W-less
W/W less and
distributed computing architectures,
applications, grids, clouds, Web services,
virtualization, dematerialization, remote
instrumentation and sensors, etc.
Share
Sh
iinfrastructure
f t t
& maximize
i i lower
l
costt
power by “following wind & sun” networks.
Sources: GENI and Inocybe
27
The “VM Turntable” Demonstrator
seamless
l
remote
rendering
Korea
KREOnet
Dynamic
g p
Lightpaths
Calgary
g y
Chicago
CA
net4
CA*net4
Starlight
VMs
APEC TEL 33,
33 Calgary,
Calgary AL,
AL Apr 24-27
24 27 2006
Live VMs migrated from Calgary to Chicago with transit through S.
Korea, resulting in just a 1.011 second of application downtime.
DRAC sets
t up and
d ttears d
down a li
lightpath
ht th w/
/ each
h migration.
i
ti
In the Blink of an Eye
> Virtual Machine teleported over thousand miles
> Seamless to external clients, w/ just a tiny ~1s glitch
> Downtime is limited despite
p high
g RTTs
– Calgary–Korea–Chicago, 1GE, RTT = 310 msec, downtime = ~1 sec
– Back to back,, 1GE,, RTT = 0.2 - 0.5 ms,, downtime = ~0.2 sec
downtime is only ~5x
while RTT is ~1,000x !!!
> Lightpath is a virtualized
> Its determinism (not the bw!)
technology
downtime
optical link
is the enabling
Thank you
> More information
> http://green-broadband.blogspot.com
htt //
b
db d bl
t
> http://free-fiber-to-the-home.blogspot.com/
>
33
