Energy Dumpster Diving - Example presentation- Paper written by: - - PowerPoint PPT Presentation

Energy Dumpster Diving

• Example presentation-

Department of Computer and Information Science (IDA) Linköping University, Sweden

Paper written by:

• M. A. Kazandjieva, B. Heller, P. Levis, C. Kozyrakis

Stanford University In Proceedings of HotPower workshop 2009 Presented by:

Simin Nadjm-Tehrani / Klervie Toczé / Rodrigo Moraes

January 27, 2020

Overview



Large computing systems



Individual elements contribute to consumption



Sources of waste difficult to identify



PowerNet sensing infrastructure



Power consumption of individual devices



Correlation with usage information



Analysis of consumption



Reveal device inefficiencies



Usage scenarios that waste energy

1

PowerNet



Large-scale distributed sensing infrastructure



Provides



Per-device energy measures



Usage statistics



Deployed in real office building environment

2

PowerNet deployment



Stanford Computer Science building



Office environment



Desktops



Monitors



Data center server rack



Small networking closet



Network switches



Some numbers



85 power meters



Utilisation data collected



15 desktops



10 servers



5 switches

3

PowerNet components

4

Wired power meters



Watt's Up .NET with Ethernet interface



Disadvantages:



Requires Ethernet port



Difficult to configure



Relatively high power consumption – 3 W



Low sampling rate – 1 Hz



High monetary cost

Similar model shown here

5

Wireless power meter



Implemented as a customised mote



Low power processor (1mA active, 1uA sleep)



Digital power meter chip



Characteristics



High rate of sampling – 14KHz



Configurable – TinyOS



No wired network required – mesh network



Lower monetary cost Mote: resource-constrained device that can sense, process, and talk wirelessly to other motes

6

Mesh network



Motes can talk to each other without cables



Data is forwarded to the sink



Mote or computer that gathers data forwarded



Sink is wired to a gateway



Gateway provides out-of-network connectivity

Image from: The Basics of Wireless Sensor Networking and its Applications http://www.ida.liu.se/~rtslab/courses/wsn/Basics.pdf

Motes Sink Computer

• r server

Gateway

7

Utilisation metering



PowerNet monitors device usage



Desktops and servers



CPU utilisation



Python script tracks utilisation



Network switches



Traffic statistics for each port



Monitor hardware counters via SNMP

 Server rack  Balanced workload

8

Data management and visualisation



Data stored in a central server



Kept in a MySQL database



Power and utilisation data correlated



Data synchronised in time using timestamps



Analysis of consumption related to activity



Data visualisation through website



Line-chart visualisation of all data



Correlated power and utilisation graphs



Searches by meter name, type, or device category

9

Case study: Desktops

10



High idle energy consumption (100 W)



Reduction



Put desktops in sleep mode when not used



User and CPU must be inactive

Case study: Desktops



Predict when it is convenient to turn machines off



Machine usage models needed



Correlation between power consumption and CPU usage

11

Case study: Monitors



Consumption comparable to desktops (40 – 130 W)



Usage pattern: almost always on, even if not in use



Consumption reduction by configuration parameters



Less brightness, less power consumption



Change desktop backgrounds (10% savings)



Total savings: 10 – 28%

Change desktop backgrounds Change of several parameters

12

Case study: Network Switches

13



Network equipment is not energy proportional



Same energy consumption independently of the usage



Maximum usage  Maximum efficiency



HP switch consumes more due to:



Fan load



Backplane structure

Case study: Server rack



10 identical 1U servers in a server rack of 40 servers



Each server consumed 245 W



But server at top of the rack consumed 20% more power



Methodology used for reasoning about odd result



Swap top and bottom servers



Top part is warmer than bottom



Same workload in all of them



Replaced server on top increased from 250 W to 270 W



Previous top server consumption back to normal 245 W

14

Case study: Server rack

 We should analyse more aspects than CPU usage alone 

Load



Temperature



Configuration

15

http://openclipart.org/detail/139525/server-rack-by-moini

Conclusions



By analysing power consumption we can:



Reduce energy consumption



Rethink system designs



Power consumption and usage pattern



Needs to be measured/recorded before further savings



Insights revealed by PowerNet:



Monitor configuration can reduce consumption (25%)



Identical server machines can have different power consumption depending on rack placement



Network equipment is not energy proportional

16

Classification

17

Design phase Production phase Use phase End-of-life phase

Resources Residues ICT Services Recycling

Energy Dumpster Diving

Discussion

18

Discussion



Which is the cost of maintenance of this system? Cost of the system itself? (not enough discussed in the paper)

19

Discussion



Which is the cost of maintenance of this system? Cost of the system itself? (not enough discussed in the paper)



Do you think this is a permanent infrastructure? Or is just to do the study? (discussion of key ideas of the paper)

20

Discussion



Which is the cost of maintenance of this system? Cost of the system itself? (not enough discussed in the paper)



Do you think this is a permanent infrastructure? Or is just to do the study? (discussion of key ideas of the paper)



Should these meters be installed in all the systems by default? (possible ideas to improve the work)

21

Discussion



Which is the cost of maintenance of this system? Cost of the system itself? (not enough discussed in the paper)



Do you think this is a permanent infrastructure? Or is just to do the study? (discussion of key ideas of the paper)



Should these meters be installed in all the systems by default? (possible ideas to improve the work)



Do you think we should create a standard to declare power values and utilisation values? (go further from the paper)

22

Discussion



Which is the cost of maintenance of this system? Cost of the system itself? (not enough discussed in the paper)



Do you think this is a permanent infrastructure? Or is just to do the study? (discussion of key ideas of the paper)



Should these meters be installed in all the systems by default? (possible ideas to improve the work)



Do you think we should create a standard to declare power values and utilisation values? (go further from the paper)



Suitable for home? (go further from the paper)

23

Web Interface



Powertron http://powernet.stanford.edu/

24

Additional material

Another similar project



The TrendMETER http://trend.polito.it

25

Additional material

Discussion



The problem of the batteries of the sensors. Which is the cost of maintenance of this system?(go further from the paper)



Do you think this is a permanent infrastructure? Or is just to do the study? (discuss the basic idea/approach of the paper)



Should these meters be installed in all the systems by default? (how to improve)



Do you think we should create a standard to distribute power values and utilisation values?



Suitable for home? (where can apply this methodology...)



Cost of the system itself?

http://www.ida.liu.se/~TDDD50

26

TinyOS



TinyOS is an operating system for resource-constrained devices



It offers you the tools to use the available features of your hardware



Written in nesC, a C dialect

May be a whole slide for TinyOS is too much because distracts the attention from the main topic

27