+ All Categories
Home > Documents > Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C....

Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C....

Date post: 01-Aug-2020
Category:
Upload: others
View: 0 times
Download: 0 times
Share this document with a friend
28
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
Transcript
Page 1: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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 UniversityIn Proceedings of HotPower workshop 2009

Presented by:Simin Nadjm-Tehrani / Klervie Toczé / Rodrigo Moraes

January 27, 2020

Page 2: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 3: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

PowerNet

Large-scale distributed sensing infrastructure

Provides Per-device energy measures Usage statistics

Deployed in real office building environment

2

Page 4: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 5: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

PowerNet components

4

Page 6: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 7: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 8: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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 Applicationshttp://www.ida.liu.se/~rtslab/courses/wsn/Basics.pdf

MotesSink

Computeror server

Gateway

7

Page 9: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 10: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 11: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 12: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Case study: Desktops

Predict when it is convenient to turn machines off Machine usage models needed Correlation between power consumption and CPU usage

11

Page 13: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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 backgroundsChange of several parameters

12

Page 14: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 15: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 16: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 17: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 18: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Classification

17

Design phase

Productionphase

Usephase

End-of-lifephase

Res

ourc

es

Res

idue

s

ICT Services

Recycling

Page 19: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Energy Dumpster Diving

Discussion

18

Page 20: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Discussion

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

19

Page 21: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 22: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 23: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 24: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 25: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Web Interface

Powertron

http://powernet.stanford.edu/24

Additional material

Page 26: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

Another similar project

The TrendMETER

http://trend.polito.it25

Additional material

Page 27: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

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

Page 28: Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C. …TDDD50/seminars/material/seminar... · 2020-01-24 · Paper written by: M. A. Kazandjieva, B. Heller, P. Levis, C.

TinyOS

TinyOS is an operating system for resource-constrained devices

It offers you the tools to use the available features ofyour hardware

Written in nesC, a C dialect

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

27


Recommended