The Efficiency Paradox and How to Save Yourself and the World

by Holly Cummins

Slide 1

Slide 1 text

No content

Slide 2

Slide 2 text

The Efficiency Paradox and How to Save Yourself and the World Holly Cummins GOTO Copenhagen

Slide 3

Slide 3 text

@holly_cummins #RedHat now senior principal software engineer helping to build Quarkus

Slide 4

Slide 4 text

@holly_cummins #RedHat now senior principal software engineer helping to build Quarkus 2007 let’s make garbage collection more efficient!

Slide 5

Slide 5 text

@holly_cummins #RedHat now senior principal software engineer helping to build Quarkus 2007 let’s make garbage collection more efficient! 2015 lean and xp makes your team more efficient!

Slide 6

Slide 6 text

@holly_cummins #RedHat now senior principal software engineer helping to build Quarkus 2007 let’s make garbage collection more efficient! 2015 lean and xp makes your team more efficient! 2022 quarkus is wonderfully efficient!

Slide 7

Slide 7 text

@holly_cummins #RedHat “this provisioning software is broken”

Slide 8

Slide 8 text

@holly_cummins #RedHat what we sold “this provisioning software is broken” 10 minute provision-time

Slide 9

Slide 9 text

@holly_cummins #RedHat what we sold “this provisioning software is broken” 10 minute provision-time 3 month provision- time what the client thought they’d got

Slide 10

Slide 10 text

@holly_cummins #RedHat what we sold “this provisioning software is broken” 10 minute provision-time 3 month provision- time what the client thought they’d got the reason 84-step pre-approval process

Slide 11

Slide 11 text

@holly_cummins #RedHat waste

Slide 12

Slide 12 text

@holly_cummins #RedHat waste is killing the planet

Slide 13

Slide 13 text

@holly_cummins #RedHat machine 1700s

Slide 14

Slide 14 text

@holly_cummins #RedHat machine 1700s energy

Slide 15

Slide 15 text

@holly_cummins #RedHat machine 1700s energy energy

Slide 16

Slide 16 text

@holly_cummins #RedHat machine 1700s energy energy useful

Slide 17

Slide 17 text

@holly_cummins #RedHat processes 1890s

Slide 18

Slide 18 text

@holly_cummins #RedHat processes time money 1890s

Slide 19

Slide 19 text

@holly_cummins #RedHat processes time money 1890s value

Slide 20

Slide 20 text

@holly_cummins #RedHat processes time money 1890s value

Slide 21

Slide 21 text

@holly_cummins #RedHat manufacturing 1900s

Slide 22

Slide 22 text

@holly_cummins #RedHat manufacturing 1900s value value

Slide 23

Slide 23 text

@holly_cummins #RedHat manufacturing 1900s value value value

Slide 24

Slide 24 text

@holly_cummins #RedHat software 1960s

Slide 25

Slide 25 text

@holly_cummins #RedHat software time electricity hardware 1960s

Slide 26

Slide 26 text

@holly_cummins #RedHat software time electricity hardware answers 1960s

Slide 27

Slide 27 text

@holly_cummins #RedHat machine energy energy useful

Slide 28

Slide 28 text

@holly_cummins #RedHat waste machine energy energy useful

Slide 29

Slide 29 text

@holly_cummins #RedHat waste is killing the planet

Slide 30

Slide 30 text

@holly_cummins #RedHat e-waste is killing the planet

Slide 31

Slide 31 text

@holly_cummins #RedHat energy waste is killing the planet

Slide 32

Slide 32 text

@holly_cummins #RedHat slow code is killing the planet

Slide 33

Slide 33 text

@holly_cummins #RedHat https://blog.linkedin.com/2017/august/3/making-linkedin-more-accessible-via-linkedin-lite

Slide 34

Slide 34 text

@holly_cummins #RedHat https://blog.linkedin.com/2017/august/3/making-linkedin-more-accessible-via-linkedin-lite “my heart sank … our new feature failed to load because of poor internet connectivity”

Slide 35

Slide 35 text

@holly_cummins #RedHat https://blog.linkedin.com/2017/august/3/making-linkedin-more-accessible-via-linkedin-lite modern web is so inefficient it is useless for part of its audience “my heart sank … our new feature failed to load because of poor internet connectivity”

Slide 36

Slide 36 text

@holly_cummins #RedHat forgotten code is killing the planet

Slide 37

Slide 37 text

#RedHat @[email protected] 25% of 16,000 servers doing no useful work https://www.anthesisgroup.com/wp-content/uploads/2019/11/Comatose-Servers-Redux-2017.pdf zombie servers

Slide 38

Slide 38 text

#RedHat @[email protected] the average server: 12 - 18% of capacity https://www.nrdc.org/sites/default/files/data-center-efficiency-assessment-IB.pdf

Slide 39

Slide 39 text

@holly_cummins #RedHat unused data is killing the planet

Slide 40

Slide 40 text

#RedHat @[email protected] 68% of data is stored and never used https://www.nrdc.org/sites/default/files/data-center-efficiency-assessment-IB.pdf https://www.seagate.com/gb/en/news/news-archive/seagates-rethink-data-report-reveals-that-68-percent-of-data-available-to-businesses-goes-unleveraged-pr-master/

Slide 41

Slide 41 text

@holly_cummins #RedHat it’s not just money

Slide 42

Slide 42 text

@holly_cummins #RedHat it’s not just electricity it’s not just money

Slide 43

Slide 43 text

@holly_cummins #RedHat it’s not just electricity it’s water it’s not just money

Slide 44

Slide 44 text

@holly_cummins #RedHat it’s not just electricity it’s water it’s e-waste it’s not just money

Slide 45

Slide 45 text

@holly_cummins #RedHat it’s not just electricity it’s water it’s e-waste it’s not just money it’s embodied carbon

Slide 46

Slide 46 text

@holly_cummins #RedHat we have solutions good news, everyone

Slide 47

Slide 47 text

@holly_cummins #RedHat making turning servers off as safe and easy as turning lights off. and then automate it.

Slide 48

Slide 48 text

@holly_cummins #RedHat solution: LightSwitchOps making turning servers off as safe and easy as turning lights off. and then automate it.

Slide 49

Slide 49 text

@holly_cummins #RedHat absurdly simple scripts my shell script to power down machines overnight saved my school €12,000

Slide 50

Slide 50 text

@holly_cummins #RedHat simple automation we used to leave our applications running all the time @darkandnerdy, Chicago DevOpsDays

Slide 51

Slide 51 text

@holly_cummins #RedHat simple automation we used to leave our applications running all the time when we scripted turning them off at night, we reduced our cloud bill by 30% @darkandnerdy, Chicago DevOpsDays

Slide 52

Slide 52 text

@holly_cummins #RedHat fancier scripts … with a frontend and a backend

Slide 53

Slide 53 text

@holly_cummins #RedHat

Slide 54

Slide 54 text

@holly_cummins #RedHat saves energy

Slide 55

Slide 55 text

@holly_cummins #RedHat saves energy saves money

Slide 56

Slide 56 text

@holly_cummins #RedHat saves energy saves money bonus: tests disaster recovery

Slide 57

Slide 57 text

@holly_cummins #RedHat saves energy saves money bonus: tests disaster recovery

Slide 58

Slide 58 text

@holly_cummins #RedHat

Slide 59

Slide 59 text

@holly_cummins #RedHat solution: faster code

Slide 60

Slide 60 text

@holly_cummins the vrroooom model

Slide 61

Slide 61 text

@holly_cummins

Slide 62

Slide 62 text

@holly_cummins invented by Dr. Vroom (really!)

Slide 63

Slide 63 text

@holly_cummins naming is the hardest problem in computer science invented by Dr. Vroom (really!)

Slide 64

Slide 64 text

@holly_cummins naming is the hardest problem in computer science

Slide 65

Slide 65 text

@holly_cummins my vrroooom model

Slide 66

Slide 66 text

No content

Slide 67

Slide 67 text

@holly_cummins #RedHat

Slide 68

Slide 68 text

@holly_cummins #RedHat these two columns are almost the same

Slide 69

Slide 69 text

@holly_cummins #RedHat Energy 1 10 100 Time 1 10 100 energy efficiency across programming languages Python Rust Java Go

Slide 70

Slide 70 text

@holly_cummins #RedHat Energy 1 10 100 Time 1 10 100 the trend line is more or less straight energy efficiency across programming languages Python Rust Java Go

Slide 71

Slide 71 text

No content

Slide 72

Slide 72 text

case study Quarkus

Slide 73

Slide 73 text

@holly_cummins #RedHat Java application frameworks were optimised for dynamism

Slide 74

Slide 74 text

@holly_cummins #RedHat Java application frameworks were optimised for dynamism dynamism has a cost

Slide 75

Slide 75 text

@holly_cummins #RedHat cloud apps are immutable now

Slide 76

Slide 76 text

@holly_cummins #RedHat cloud apps are immutable now

Slide 77

Slide 77 text

@holly_cummins #RedHat a highly dynamic runtime in a container is pointless

Slide 78

Slide 78 text

@holly_cummins Java dynamism

Slide 79

Slide 79 text

@holly_cummins Java dynamism build time

Slide 80

Slide 80 text

@holly_cummins Java dynamism build time runtime

Slide 81

Slide 81 text

@holly_cummins Java dynamism build time runtime

Slide 82

Slide 82 text

@holly_cummins Java dynamism packaging (maven, gradle…) build time runtime

Slide 83

Slide 83 text

@holly_cummins Java dynamism build time runtime

Slide 84

Slide 84 text

@holly_cummins Java dynamism build time runtime

Slide 85

Slide 85 text

@holly_cummins Java dynamism > build time runtime load and parse • config files • properties • yaml • xml • etc.

Slide 86

Slide 86 text

@holly_cummins Java dynamism > build time runtime

Slide 87

Slide 87 text

@holly_cummins Java dynamism @ @ > build time runtime • classpath scanning and annotation discovery • attempt to load class to enable/disable features

Slide 88

Slide 88 text

@holly_cummins Java dynamism @ @ > build time runtime

Slide 89

Slide 89 text

@holly_cummins Java dynamism @ @ > build time runtime build a metamodel of the world

Slide 90

Slide 90 text

@holly_cummins Java dynamism @ @ > build time runtime

Slide 91

Slide 91 text

@holly_cummins Java dynamism @ @ > build time runtime start • thread pools • I/O • etc.

Slide 92

Slide 92 text

@holly_cummins Java dynamism @ @ > build time runtime ready to do work!

Slide 93

Slide 93 text

@holly_cummins what if we start the application more than once? @ @ >

Slide 94

Slide 94 text

@holly_cummins what if we start the application more than once? @ @ > @ @ >

Slide 95

Slide 95 text

@holly_cummins what if we start the application more than once? @ @ > @ @ > @ @ >

Slide 96

Slide 96 text

@holly_cummins what if we start the application more than once? @ @ > @ @ > @ @ > @ @ >

Slide 97

Slide 97 text

@holly_cummins what if we start the application more than once? @ @ > @ @ > @ @ > @ @ >

Slide 98

Slide 98 text

@holly_cummins what if we start the application more than once? @ @ > @ @ > @ @ > @ @ > so much work gets redone every time

Slide 99

Slide 99 text

@holly_cummins Hibernate speed example: JTA auto-wiring

Slide 100

Slide 100 text

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”);

Slide 101

Slide 101 text

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”);

Slide 102

Slide 102 text

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”);

Slide 103

Slide 103 text

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”); Class.forName(“AnotherJTAImplementation”);

Slide 104

Slide 104 text

Slide 105

Slide 105 text

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”); Class.forName(“AnotherJTAImplementation”); Class.forName(“NicheJTAImplementation”); …

Slide 106

Slide 106 text

Slide 107

Slide 107 text

Slide 108

Slide 108 text

Slide 109

Slide 109 text

Slide 110

Slide 110 text

@holly_cummins it’s not just JTA this happens for lots of internal service bindings

Slide 111

Slide 111 text

@holly_cummins the true cost of loaded classes isn’t just memory + start time

Slide 112

Slide 112 text

@holly_cummins the true cost of loaded classes isn’t just memory + start time method dispatching:

Slide 113

Slide 113 text

@holly_cummins interface the true cost of loaded classes isn’t just memory + start time method dispatching:

Slide 114

Slide 114 text

@holly_cummins unused implementation the one we want interface unused implementation unused implementation the true cost of loaded classes isn’t just memory + start time method dispatching:

Slide 115

Slide 115 text

@holly_cummins unused implementation the one we want interface unused implementation unused implementation the true cost of loaded classes isn’t just memory + start time method dispatching:

Slide 116

Slide 116 text

@holly_cummins unused implementation the one we want interface megamorphic call slow dispatching unused implementation unused implementation the true cost of loaded classes isn’t just memory + start time method dispatching:

Slide 117

Slide 117 text

@holly_cummins the true cost of loaded classes isn’t just memory + start time the one we want interface

Slide 118

Slide 118 text

@holly_cummins the true cost of loaded classes isn’t just memory + start time the one we want monomorphic call fast dispatching interface

Slide 119

Slide 119 text

@holly_cummins how do we fix all this?

Slide 120

Slide 120 text

@holly_cummins @ @ > build time runtime what if we initialize at build time?

Slide 121

Slide 121 text

@holly_cummins @ @ > build time runtime what if we initialize at build time?

Slide 122

Slide 122 text

@holly_cummins @ @ > build time runtime start • thread pools • I/O • etc. what if we initialize at build time?

Slide 123

Slide 123 text

@holly_cummins @ @ > build time runtime ready to do work! start • thread pools • I/O • etc. what if we initialize at build time?