Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI,...
-
Upload
egbert-houston -
Category
Documents
-
view
215 -
download
0
Transcript of Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI,...
![Page 1: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/1.jpg)
Advanced Energy Management
in Cloud Computing multi data center environments
Giuliana Carello, DEI, Politecnico di [email protected]
Danilo Ardagna, Antonio CaponePolitenico di Milano, 12 June 2012
![Page 2: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/2.jpg)
CLOUD CONSUMPTION
About 0.5% of global electric power consumption is due to Data Centers (DC)
In developed country: ‐UK: 2.2-3.3% ‐USA: 1.5%
From the environmental point of view:‐2% of global CO2 emissions
2005
2020
0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
% European IT consumption
Cellular phone NetworkTelecom NetworkServer & Data Center
Source: EU Commission
![Page 3: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/3.jpg)
ENVIRONMENTAL IMPACT
High consumption, environmental impact and energy costs of Cloud Computing
A way out: “GREEN” CLOUD COMPUTING
New Servers costsEnergy and cooling costs
IT costsNumber of Servers (M units)
…and costsEnvironmental impact
Source: EU Commission
![Page 4: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/4.jpg)
OUR APPROACH
Green Cloud project: Innovative strategies for resource allocation in Cloud Computing multi data centers
systems Exploiting different energy costs and different workload due to geographically
distributed data centers Integrated Data Center and network management Green energy sources utilization
Definition of Linear Programming Optimization models Scenario definition and analysis
![Page 5: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/5.jpg)
SCENARIO
Optimization over 24 hours for each time band Set of geographically distributed DC Traffic profile for each Data Center over the day Full-connected network Possibility to redirect requests from one DC to another Energy cost for both DC and network: fixed + linear
NETWORK
Request forwarding is due to: Lower energy cost Tradeoff between DC and network costs Capacity limits not reached
![Page 6: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/6.jpg)
BROWN MODEL
DECISION VARIABLES:
Requests rate of type k incoming to DC i served in DC j with a type l server
OBJECTIVE FUNCTION:
Minimization of the costs over 24 hours Cost for operating servers Cost for switching on/off servers Bandwidth utilization Number of active link and link switching costs
![Page 7: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/7.jpg)
BROWN MODEL
CONSTRAINTS:DATA CENTERS
The whole incoming traffic has to be served Each request must be served by suitable server DCs can handle a finite number of requests Server have a utilization limit
![Page 8: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/8.jpg)
BROWN MODEL
CONSTRAINTS:DATA CENTERS
The whole incoming traffic has to be served Each request must be served by suitable server DCs can handle a finite number of requests Server have a utilization limit
NETWORKS Links have limited bandwidth
![Page 9: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/9.jpg)
BROWN MODEL
CONSTRAINTS:DATA CENTERS
The whole incoming traffic has to be served Each request must be served by suitable server DCs can handle a finite number of requests Server have a utilization limit
NETWORKS Links have limited bandwidthSWITCHING
Implemented in AMPL language using IBM ILOG CPLEX 12.1 solver
![Page 10: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/10.jpg)
GREEN MODEL
ASSUMPTIONS: DCs can be powered through renewable sources:‐ Solar‐ Geothermic‐ Wind
Limited amount of green energy available Only autonomous production Minor cost with respect to brown energy sources
IMPLICATIONS: Green powered DCs are preferred CO2 reduction
![Page 11: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/11.jpg)
DATA CENTER MAP
Data Center locations
![Page 12: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/12.jpg)
INCOMING TRAFFIC
Starting from the basic profile: Re-scaling according to the geographical location Temporal shift to consider time zone differences
Analysis of the model behavior with respect to a growing number of incoming requests:2 – 3 – 16 – 30 – 40 billions of daily requests
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240
1
Time band
![Page 13: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/13.jpg)
ENERGY COST
Differential trend according to: Energy market studies Geographical location Time band
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240
10
20
30
40
50
60
70
West USA
East USA
Europe
Asia
Time band
€/M
Wh
![Page 14: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/14.jpg)
GREEN DATA CENTERS
Location derived from green energy sources availability
![Page 15: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/15.jpg)
GREEN ENERGY AVAILABILITY
Different trends and energy production according to the different green energy sources
Analysis based on the dimension and location of the DCs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240
50
100
150
200
250
Mountain View, CA - Solar power
Seattle, WA - Geothermal power
Frankfurt, GER - Wind power
Sao Paolo, BRA - Solar power
Time bands
kWh
![Page 16: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/16.jpg)
REQUESTS REDIRECTION
Requests are redirected in the macro-area where the DC energy cost is lower
0
50
100
150
200
250
300
350
05
10152025303540
0
100
200
300
400
500
600
05
1015202530354045
0
20
40
60
80
100
120
140
0
10
20
30
40
50
60
70
050
100150200250300350400450500
0
10
20
30
40
50
60
70
West US East US Europe Asia
Activ
e se
rver
s€/
MW
h
![Page 17: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/17.jpg)
POLICIES TO BE COMPARED
Brown Model: Possibility of request redirection from one DC
to another
Green Model: Extension of the Brown Model taking into
account green energy sources
Base Case: Requests execution in local
![Page 18: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/18.jpg)
EXPENSES REDUCTION
2 3 16 30 400
2000
4000
6000
8000
10000
12000
14000
16000
35% 39%
39%
37%
Base
Brown
Billions of daily requests
Ener
gy e
xpen
ses
(€/d
ay)
Unf
easi
ble
solu
tion
BROWN MODEL
![Page 19: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/19.jpg)
2 3 16 30 400
50,000
100,000
150,000
200,000
250,000
300,000
350,000
400,000
450,000
500,000
12% 7%
9%
10%
BaseBrown
Billions of daily requests
Ener
gy c
onsu
mpti
on (k
Wh/
day)
Unf
easi
ble
solu
tion
ENERGY CONSUMPTION
Model priority: minimizing expenses Higher energy consumption due to network utilization for request forwarding
![Page 20: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/20.jpg)
2 3 16 30 400
2000
4000
6000
8000
10000
12000
14000
56% 49%
43%
40%
Base
Green
Billions of daily requests
Ener
gy e
xpen
ses
(€/d
ay)
Unf
easi
ble
solu
tion
EXPENSES REDUCTION
GREEN MODEL
![Page 21: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/21.jpg)
ENERGY CONSUMPTION
2 3 16 30 40 -
100,000
200,000
300,000
400,000
500,000
600,000
DC Green
DC Brown
Network
Billions of daily requests
Ener
gy c
onsu
mpti
on (k
Wh)
Green energy saturation Possibility to better exploit the requests redirection by investing in
green energy sources
![Page 22: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/22.jpg)
CO2 EMISSIONS
2 3 16 30 400
50
100
150
200
250
300
350
400
57% 41%
11%
8%
7%BrownGreen
Billions of daily requests
CO2
emis
sion
s (t
ons/
day)
Comparison of brown and green models based on the CO2 emissions Maximum amount of green energy availability fixed at 8%
![Page 23: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/23.jpg)
MODEL FOOTPRINT Reduction of greenhouse gases according to
different green energy availability
CO2 EMISSIONS
![Page 24: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/24.jpg)
MODEL GAP
Two hours time limit, average gap 5%
![Page 25: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/25.jpg)
MODEL SOLVING TIME
![Page 26: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/26.jpg)
FUTURE DEVELOPMENT
The model proved to be: Robust, flexible and scalable
POSSIBLE EXTENSIONS: Non fully connected network topology SLA considerations and response time Tests on larger instances
![Page 27: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/27.jpg)
CONCLUSIONS
ECONOMICAL POINT OF VIEW: Optimization of Network and DCs jointly Expenses reduction up to 40%
ENVIRONMENTAL POINT OF VIEW: Reduction of greenhouse gas emissions Optimizing costs linked to the Carbon Credit system Promoting the use of green energy sources
![Page 28: Advanced Energy Management in Cloud Computing multi data center environments Giuliana Carello, DEI, Politecnico di Milano carello@elet.polimi.it Danilo.](https://reader036.fdocuments.us/reader036/viewer/2022062518/56649e295503460f94b170ae/html5/thumbnails/28.jpg)
THANK YOU FOR THE
ATTENTION!!!
ANY QUESTIONS?