Can Cloud Computing
and Optical Networking
S
Save
th
the Planet?
Pl
t?
Rod Tucker, Jayant Baliga, Robert Ayre, Kerry Hinton
ARC Special Research Centre for
Ultra-Broadband Information Networks (CUBIN)
University of Melbourne
r.tucker@ee.unimelb.edu.au
Cloud Computing
What do the following have in common?
• Happiness
• Love
• Marketing Hype
• Recession
• Cloud Computing
p
g
They all mean different things to different people
Summary
Cloud Computing as a Logistics Problem
Key elements of logistics
• Processing
• Storage
• Transport
Energy costs
Cloud Computing Energy Model
Computer,
Storage
Enterprise Data Center
Storage
Servers,
Processors
PON Access Network
Fiber
OLT
Core Network
Core
Router
Broadband
Network
G
Gateways
OXC
Edge
Routers
Ethernet
Switch
Servers,
Processors
M t /Ed Network
Metro/Edge
N t
k
Storage
Public Data Center
Cloud
Splitter
ONU
Total Energy
Consumption =
Energy/bit used in
• Storage
• Transport
• Processing
X
Total number of bits
Network Energy Model
1000
2009
Energy per bit (n
nJ)
100
10
1
01
0.1
0 01
0.01
Optical
Amp
PoS Ethernet Core
PON
Server
WDM Switch Router ONU
(10 Mb/s)
Tx/Rx
Data Centre Energy Model
Edge Router
26 nJ/b
Servers
430 nJ/b
Storage
8 pW/b
p
[29 nJ/hr/b]
LAN – 2 stages of aggregation
using
g Ethernet switches
Baliga et al., OThT6
Watts not Joules
User Hardware Energy Model
• End user computer:
– High Energy Use Desktop 140 Watts
• Eg. Intel Quad-core, or older P4, mid-range graphics
– Modern mid-range Desktop 70 Watts
• Eg. Intel Core Duo 2.5 GHz, 2 GB RAM, mid-range graphics
– Low-end Desktop/Thin Client 18 Watts
• Eg.
g Intel Atom 1.6 GHz,, 2 GB RAM,, chipset-based
p
graphics
g
p
– Ignore monitor (LCD/CRT) energy consumption
• This will be similar in all our comparisons
Outsource to Cloud or Oversize Computer?
Should I use a low-end Computer & Outsource Jobs
Computing
Resource
Service
Center
Metro/Edge
Network
Outsource jobs
Access
Network
Low energy
consumption.
- OR Use a high-end
g
computer
p
that
can handle any of my jobs
But high energy consumption even
for light jobs
Consider Three Scenarios
1. Software-as-a-Service
– Stored on user’s computer with updates downloaded regularly
• Eg current anti-virus software model extended to mainstream
applications, MS Update (weekly patch) model
2 S
2.
Service
i B
Bureau
– Most tasks done on lower end user machine,
outsource the “big” jobs
3. Computing-as-a-Service
– Hosted and run on provider’s
provider s computer “farm”
farm with data initially
uploaded from user
• User data may be hosted or locally stored
• Like enterprise “Thin
Thin Client”
Client model
Can cloud computing save energy ?
1. SoftwareSoftware-asas-a-Service
Data
Center
Service
Center
Metro/Edge
Network
Software as a Service
Software-as-a-Service
• Software maintenance
I want to use
the Word
Processor
Access
Network
Mid-range
Computer
1. SoftwareSoftware-asas-a-Service
Data
Center
Service
Center
Here is the latest
version or patch
Mid-range
Computer
Metro/Edge
Network
Software as a Service
Software-as-a-Service
• Software maintenance
Access
Network
2. Service Bureau
Computing
Resource
Service
Center
Here is my raw
video footage
and edit
commands
Mid-range
Computer
Metro/Edge
Network
Service Bureau – Hardware & Software
•Service Bureau for “Big” Jobs
Access
Network
2. Service Bureau
Computing
Resource
Service
Center
Here is your
DVD Image
Mid-range
Computer
Metro/Edge
Network
Service Bureau – Hardware & Software
•Service Bureau for “Big” Jobs
Access
Network
3. Computing
Computing--as
as--a-Service
Enterprise
Server
Service
Center
Metro/Edge
Network
Computing (Hardware and Software-asa-Service)
•Centralised Storage and Processing
Access
Network
Here is the new
data for my
spreadsheet
p
Low-end or
“Thin Client”
C
Computer
t
3. ComputingComputing-asas-a-Service
Enterprise
Server
Service
Center
Metro/Edge
Network
Computing (Hardware and Software-asa-Service)
•Centralised Storage and Processing
Access
Network
Here is the
output file to
print or e-mail
p
Low-end or
“Thin Client”
C
Computer
t
1. SoftwareSoftware-asas-a-Service Example
Store on Local Disk or in the Cloud?
•
Energy
gy impact
p
of the
“Software as a Service” &
“Storage as a Service” models
•
Compare the energy cost to
store & transport 50 MB
bl k to
blocks
t the
th user,
cf. store on local drive
•
Data
Centre
Service
Centre
Metro
Network
U
Users
require
i 1GB and
d 5GB “cloud”
“ l d” storage
t
Access
Network
1. SoftwareSoftware-asas-a-Service Example
Store on Local Disk or in the Cloud?
Storage of application & data “in the cloud” compared with storing on a local disk.
50 MBytes per download. Modern laptop-style
laptop style HDD 20% read/write and 80% idle.
Power C
Consump
ption (W))
100
Transport
energy dominates
10
L
Local
l HDD
1
0.1
1 GB Cloud Storage
0.01
0.01
0.1
5 GB Cloud Storage
1
D
Downloads
l d per Hour
H
10
100
2. Service Bureau Example
Computer used for 20 hrs/week, plus some video encoding of ½ hour videos
Average P
A
Power Con
nsumptio
on, W
180
Older Computer
160
Transport
energy dominates
140
120
Mid-range Computer
100
80
60
40
Low-end Computer & outsourcing
20
0
0
20
40
60
80
Number of Encodings per Week
100
120
3. Computing
Computing--asas-a-Service Example
Thin Client or Small PC in Office?
•
Office situation
– non-intensive tasks,
•
Enterprise
Server
Service
Center
Should this office deploy:
– a low-end machine per user
OR “dumb, thin clients”?
Metro/Edge
Network
•
Commonly, thin client would receive and display
HTML formatted data from the server
•
Thin client
Thi
li t capability
bilit may be
b limited
li it d to
t displaying
di l i
bitmaps developed from its server
Access
Network
3. Computing
Computing--asas-a-Service Example
Pow
wer Saving
g using T
Thin Clien
nt (W)
Thin Client compared to Small PC in Office
7.5
“local” server
5
2.5
Distant Server
0
-2.5
1%
2%
5%
10%
-5
0
1
2
3
4
Average Bitrate for Screen Refresh Mb/s
Thin client solution “green”
green only when the server is local
5
% of screen
area updated
per second
Summary
• Cloud Storage and Cloud Computing can provide
energy savings
– More so for less-demanding users
– For handling the occasional intensive task
• But...
– Advantages may be lost when data transfer is large
– This is where optical networking needs to come to the
rescue
• Cloud
Cl d C
Computing
ti may nott save the
th planet….
l
t
– But if used appropriately, it could help to mitigate the
growth in energy
g
gy consumption
p
in our industry
y
– A classical Logistics problem