g Green Mining and the Perils of Mining Energy Profjles Abram - - PowerPoint PPT Presentation

▶
g
SMART_READER_LITE
LIVE PREVIEW

g Green Mining and the Perils of Mining Energy Profjles Abram - - PowerPoint PPT Presentation

Jun 19, 2023 222 likes •754 views

g Green Mining and the Perils of Mining Energy Profjles Abram Hindle (hindle1@ualberta.ca) Department of Computing Science University of Alberta http://softwareprocess.es/ with Karan Aggarawal*, Candy Pang*, Chenlei Zhang*, Kent Rasmussen*,

slide-1
SLIDE 1

g

slide-2
SLIDE 2

Abram Hindle (hindle1@ualberta.ca) Department of Computing Science University of Alberta http://softwareprocess.es/ with Karan Aggarawal*, Candy Pang*, Chenlei Zhang*, Kent Rasmussen*, Joshua Campbell*, Alex Wilson*, Jed Barlow*, Stephen Romansky*, w/ Eleni Stroulia, Bram Adams, Ahmed Hassan, Daniel German

Green Mining and the Perils of Mining Energy Profjles

slide-3
SLIDE 3
slide-4
SLIDE 4

energy

slide-5
SLIDE 5

Programmers are responsible for Software Power Use!

slide-6
SLIDE 6

In general, programmers lack knowledge about software energy consumption, but they are more knowledgeable about software energy consumption on mobile devices than on desktop computers.

[Pang et al.] http://webdocs.cs.ualberta.ca/~hindle1/2014/green-programmers/

slide-7
SLIDE 7

MOV [bufgerloc], EAX

slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10

Energy Profjle of Single Versions

slide-11
SLIDE 11

Android Firefox Energy Profile

Android Fennec Revision Mean Watts 518a79e90444 3f0b94325b80 7d4268f8884c bfd85c9652fa fef5f202b2dc 0.6 0.8 1.0 1.2 1.4

Mean Watts Software Revision 0.6 0.8 1.0 1.2 1.4 518a79e90444 7d4268f8884c fef5f202b2dc

slide-12
SLIDE 12

Android Calculator Energy Profile

slide-13
SLIDE 13

Energy Profjle: Same test, Same Device

  

1 5 1 6 1 7 B

J

  • u

l e s

a p p _

1 _ S u i t e 1 m s g J

  • u

l e s

slide-14
SLIDE 14

Energy Profjle: Difgerent device, difgerent response

       

1 5 1 7 5 A B C D

A p p V e r s i

  • n

s

J

  • u

l e s

1 _ S u i t e 1 m s g J

  • u

l e s

slide-15
SLIDE 15

Error rors Tha That s silver er ba band is t nd is the he manuf ufacturing cturing err error t r toler

  • leranc

nce

slide-16
SLIDE 16

Error rors Mult ultiple i le ins nstan ances ces of t f the e same m ame model o

  • del of ph

f phone ne wi will pro ll produce di duce difger fgerent m t meas easurem urements ts du due t e to e

  • erro

rror t r tolera

  • lerances

nces i in Man anufa ufactu cturi ring ng! This i impli mplies es a n a nee eed for for nor normali malizat ation n and rep nd repeated t ed tes ests ts!

slide-17
SLIDE 17
slide-18
SLIDE 18

Phone Ras aspberr rry P y Pi Power Sup r Supply Ardui uino USB H USB Hub/P /PS

slide-19
SLIDE 19

Elaborate Job Queue

[MSR 2014]

slide-20
SLIDE 20

How does it work?

slide-21
SLIDE 21

Fake the Battery

slide-22
SLIDE 22

Control the USB Connection

A r d u i n

  • T

I P 1 2 7

USB Phone USB Computer D i g i t a l 4 G N D

1 k

slide-23
SLIDE 23

Row 1 Row 2 Row 3 Row 4 2 4 6 8 10 12 Column 1 Column 2 Column 3

Timeseries graph and Component Breakdown

slide-24
SLIDE 24

Aggregate Plotting How does my energy profjle change

  • ver time?
slide-25
SLIDE 25

Aggregate by Component and Version! How does power use difger across scenarios and versions?

slide-26
SLIDE 26

Green Miner

This colour is what happens When you don't specify UI Look and feel requirements. [MSR 2014]

slide-27
SLIDE 27

Per Component Breakdown

slide-28
SLIDE 28

Sometimes we need to debug the tests with some context

1

slide-29
SLIDE 29

Green Miner is an enabler

  • Continuous Integration and Testing for Energy

Regressions

  • Repeatable, logged, measureable framework.
  • Enables asking of questions and repeating

experiments.

  • Some examples to follow
slide-30
SLIDE 30
slide-31
SLIDE 31

Does Adblocking Save Power?

[GREENS 2014] MO MORE Adblock Saves power Versus Nothing

  • r very large

/etc/hosts files

slide-32
SLIDE 32

Does Adblocking Save Power?

[GREENS 2014] MO MORE Adblock Saves power Versus Nothing

  • r very large

/etc/hosts files

Adblocking saved 0.01 to 0.06W per test (~5J every 100 seconds)

slide-33
SLIDE 33

Which UI Theme would use More Power?

White on Black UI Theme Black on White UI Theme

slide-34
SLIDE 34
slide-35
SLIDE 35

What Causes Software Power Use?

[Aggarwal et al.]

slide-36
SLIDE 36

What causes Software Power Use?

[Aggarwal et al.]

Syscalls matter but Syscalls can always Vary per run of the Same test!

slide-37
SLIDE 37

What causes Software Power Use?

[Aggarwal et al.]

Syscalls matter but Syscalls can always Vary per run of the Same test!

If the system call profjle changes signifjcantly from the previous version, it is probable (~50%) that the application’s energy consumption profjle has changed as well.

slide-38
SLIDE 38

What causes Software Power Use?

  • Rule of thumb model:

If a cumulative count

  • f a syscall changes

signfjcantly between versions there's a good chance of a signfjciant change in power use and energy consumption!

[Aggarwal et al.]

slide-39
SLIDE 39

Does logging matter?

[Di Nardio et al.]

Bufgering hides, delays and defers The cost of IO

slide-40
SLIDE 40

Does logging matter?

. 8 . 9 1 .

M e a n Wa t t s

_ Wa i t f

  • r

w a t t l

  • g1

_ S u i t e 1 m s g 2 _ S u i t e 1 m s g 3 _ S u i t e 1 m s g 4 _ S u i t e 1 m s g 5 _ S u i t e 1 m s g 6 _ S u i t e _ 1 m s g 7 _ I d l e t i m e

A p p V e r s i

  • n

s

[Di Nardio et al.]

Bufgering hides, delays and defers The cost of IO

slide-41
SLIDE 41

Methodological Bottleneck Applied Exhaustively to all revisions Revisions are measured multiple times for stability

Android Fennec Revision Mean Watts 518a79e90444 3f0b94325b80 7d4268f8884c bfd85c9652fa fef5f202b2dc 0.6 0.8 1.0 1.2 1.4
slide-42
SLIDE 42

The Research Hurdle: Measuring every revision takes too long!

slide-43
SLIDE 43

Pareto frontier: Most cost effective approximations per time spent

slide-44
SLIDE 44

Pareto frontier: Most cost effective approximations per time spent

Mo More r revisions ions t tested Hig igher fj r fjnal ac l accu curacy Le Less r revision ions t tested Lo Lower fj fjnal a al accu curac acy

slide-45
SLIDE 45

Pareto frontier: Most cost effective approximations per time spent

Half o lf of t f the he w work rk Sm Small Di ll Difg fgerence i in RMSE n RMSE

slide-46
SLIDE 46

Pareto frontier: Most cost effective approximations per time spent

Half o lf of t f the he w work rk Sm Small Di ll Difg fgerence i in RMSE n RMSE

slide-47
SLIDE 47

Pareto frontier: Most cost effective approximations per time spent

A t thir ird o

  • f t

f the he w work

  • rk

Sm Small Di ll Difg fgerence i in RMSE n RMSE

slide-48
SLIDE 48

Pareto frontier: Most cost effective approximations per time spent

1/3 /3 t to 1 1/2 /20th of

  • f t

the W Work rk Larger Di r Difg fgerence in in RMSE RMSE

slide-49
SLIDE 49

Pareto frontier: Most cost effective approximations per time spent

Tim ime & Accu curac acy

slide-50
SLIDE 50

Green Mining Needs You Got a question about power? Looking for collaborators. Green Miner is usable remotely.

slide-51
SLIDE 51

Conclusions

1 measurement is not enough Energy profjles change Measurement Granularity Software evolves, and forks,

  • - it's not always now.

Green Miner is available For collaboration!

http://greenmining.softwareprocess.es/