Slide 1
Slide 1 text
is efficiency
a good
thing?
holly cummins
7 december, 2023
tech.rocks
Slide 2
Slide 2 text
@holly_cummins
#RedHat
now
senior principal software engineer
helping to build Quarkus
Slide 3
Slide 3 text
@holly_cummins
#RedHat
now
senior principal software engineer
helping to build Quarkus
2007
let’s make garbage
collection more
efficient!
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!
2015
lean and xp makes
your team 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!
2022
quarkus is
wonderfully efficient!
Slide 6
Slide 6 text
@holly_cummins #RedHat
“j’aimerais bien être inefficace”
Slide 7
Slide 7 text
@holly_cummins #RedHat
“j’aimerais bien être inefficace”
none us of us say this
Slide 8
Slide 8 text
@holly_cummins #RedHat
Slide 9
Slide 9 text
@holly_cummins #RedHat
Slide 10
Slide 10 text
@holly_cummins #RedHat
machine
1700s
Slide 11
Slide 11 text
@holly_cummins #RedHat
machine
1700s
energy
Slide 12
Slide 12 text
@holly_cummins #RedHat
machine
1700s
energy energy
Slide 13
Slide 13 text
@holly_cummins #RedHat
machine
1700s
energy energy
useful
Slide 14
Slide 14 text
@holly_cummins #RedHat
processes
1890s
Slide 15
Slide 15 text
@holly_cummins #RedHat
processes
time
money
1890s
Slide 16
Slide 16 text
@holly_cummins #RedHat
processes
time
money
1890s
value
Slide 17
Slide 17 text
@holly_cummins #RedHat
processes
time
money
1890s
“interaction at scale”
value
Slide 18
Slide 18 text
@holly_cummins #RedHat
manufacturing
1900s
Slide 19
Slide 19 text
@holly_cummins #RedHat
manufacturing
1900s
value
value
Slide 20
Slide 20 text
@holly_cummins #RedHat
manufacturing
1900s
value
value
value
lean
Slide 21
Slide 21 text
@holly_cummins #RedHat
software
1960s
Slide 22
Slide 22 text
@holly_cummins #RedHat
software
time
electricity
hardware
1960s
Slide 23
Slide 23 text
@holly_cummins #RedHat
software
time
electricity
hardware
answers
1960s
Slide 24
Slide 24 text
@holly_cummins #RedHat
takeaway
consider all kinds of efficiency
-processes
-production
-software
Slide 25
Slide 25 text
@holly_cummins #RedHat
after 200 years, we’re getting good at efficiency
Slide 26
Slide 26 text
@holly_cummins #RedHat
source: http://geosci.uchicago.edu/~moyer/GEOS24705/2016/Assignments/PS7.pdf
steam engine
efficiency
Slide 27
Slide 27 text
@holly_cummins #RedHat
koomey’s law: the energy efficiency of
computers doubles roughly every 18
months
computer
efficiency
Slide 28
Slide 28 text
@holly_cummins #RedHat
koomey’s law: the energy efficiency of
computers doubles roughly every 18
months
computer
efficiency
Slide 29
Slide 29 text
software
efficiency
Slide 30
Slide 30 text
starting a java application in 2010
Slide 31
Slide 31 text
starting a java application in 2010
slower than making tea
Slide 32
Slide 32 text
starting a java application in 2023
Slide 33
Slide 33 text
starting a java application in 2023
faster than a lightbulb
Slide 34
Slide 34 text
starting a java application in 2023
faster than a lightbulb
quarkus
Slide 36
Slide 36 text
@holly_cummins #RedHat
we have automated all the things
Slide 37
Slide 37 text
@holly_cummins #RedHat
infrastructure as code
Slide 38
Slide 38 text
@holly_cummins #RedHat
… us as code
Slide 39
Slide 39 text
@holly_cummins #RedHat
Slide 40
Slide 40 text
@holly_cummins #RedHat
Slide 41
Slide 41 text
@holly_cummins #RedHat
Slide 42
Slide 42 text
@holly_cummins #RedHat
@holly_cummins
Slide 43
Slide 43 text
@holly_cummins #RedHat
@holly_cummins
Slide 44
Slide 44 text
@holly_cummins #RedHat
@holly_cummins
all my cool ideas were
apparently not very original!
Slide 45
Slide 45 text
@holly_cummins #RedHat
even after 200 years, we are not
always very good at efficiency
but.
Slide 46
Slide 46 text
@holly_cummins #RedHat
process inefficiency
Slide 47
Slide 47 text
@holly_cummins #RedHat
“this provisioning
software is broken”
Slide 48
Slide 48 text
@holly_cummins #RedHat
what we sold
“this provisioning
software is broken”
10 minute
provision-time
Slide 49
Slide 49 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 50
Slide 50 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 51
Slide 51 text
@holly_cummins #RedHat
software inefficiency
Slide 52
Slide 52 text
@holly_cummins #RedHat
https://blog.linkedin.com/2017/august/3/making-linkedin-more-accessible-via-linkedin-lite
Slide 53
Slide 53 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 54
Slide 54 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 56
Slide 56 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 57
Slide 57 text
#RedHat
@[email protected]
zombie
“they haven't delivered any information or
computing services for six months or more”
Slide 58
Slide 58 text
#RedHat
@[email protected]
29%
of 4,000 active less
than 5% of the time
https://www.anthesisgroup.com/wp-content/uploads/2019/11/Comatose-Servers-Redux-2017.pdf
under-utilised servers
Slide 59
Slide 59 text
#RedHat
@[email protected]
the average server:
12 - 18% of capacity
30 - 60 % of maximum power
https://www.nrdc.org/sites/default/files/data-center-efficiency-assessment-IB.pdf
Slide 60
Slide 60 text
@holly_cummins #RedHat
it’s not just money
Slide 61
Slide 61 text
@holly_cummins #RedHat
it’s not just electricity
it’s not just money
Slide 62
Slide 62 text
@holly_cummins #RedHat
it’s not just electricity
it’s water
it’s not just money
Slide 63
Slide 63 text
@holly_cummins #RedHat
it’s not just electricity
it’s water
it’s e-waste
it’s not just money
Slide 64
Slide 64 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 65
Slide 65 text
@holly_cummins #RedHat
we cling on to inefficiency
Slide 66
Slide 66 text
@holly_cummins #RedHat
true story from the software marketing trenches
Slide 67
Slide 67 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
true story from the software marketing trenches
Slide 68
Slide 68 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
ooh, no, definitely
don’t want that
true story from the software marketing trenches
Slide 69
Slide 69 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
ooh, no, definitely
don’t want that
and it’s yours for only –
wait, what?
true story from the software marketing trenches
Slide 70
Slide 70 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
ooh, no, definitely
don’t want that
and it’s yours for only –
wait, what? if we became more
efficient our headcount
would be reduced
true story from the software marketing trenches
Slide 71
Slide 71 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
ooh, no, definitely
don’t want that
and it’s yours for only –
wait, what? if we became more
efficient our headcount
would be reduced
my status
would be lower
true story from the software marketing trenches
Slide 72
Slide 72 text
@holly_cummins #RedHat
our amazing product makes
your team more efficient
ooh, no, definitely
don’t want that
and it’s yours for only –
wait, what? if we became more
efficient our headcount
would be reduced
my status
would be lower
true story from the software marketing trenches
…
Slide 73
Slide 73 text
@holly_cummins #RedHat
if I tell my boss about an
inefficiency, what if my friends get laid
off because less staff are needed after
it’s fixed?
Slide 74
Slide 74 text
@holly_cummins #RedHat
takeaway
drive out inefficiency
Slide 75
Slide 75 text
@holly_cummins #RedHat
takeaway
drive out inefficiency
⁃ stupid processes
⁃ bloated software
⁃ terrible resource utilisation
Slide 76
Slide 76 text
@holly_cummins #RedHat
takeaway
drive out inefficiency
⁃ stupid processes
⁃ bloated software
⁃ terrible resource utilisation
make sure incentive structures do not reward inefficiency
Slide 77
Slide 77 text
@holly_cummins #RedHat
once you’ve eliminated the obvious waste …
go further
Slide 78
Slide 78 text
@holly_cummins #RedHat
“There is nothing so useless as
doing efficiently that which
should not be done at all."
Peter Drucker
Slide 79
Slide 79 text
@holly_cummins #RedHat
defect triage team
Slide 80
Slide 80 text
@holly_cummins #RedHat
defect triage team
Slide 81
Slide 81 text
@holly_cummins #RedHat
let’s … set the priority
to ‘medium’!
defect triage team
Slide 82
Slide 82 text
@holly_cummins #RedHat
let’s … set the priority
to ‘medium’!
defect triage team
could we turn this
whole meeting into a
small shell script?
Slide 83
Slide 83 text
@holly_cummins #RedHat
let’s … set the priority
to ‘medium’!
defect triage team
could we turn this
whole meeting into a
small shell script?
what if … we just got rid
of the priority field?
Slide 84
Slide 84 text
@holly_cummins #RedHat
instead of using ai to fill in boilerplate,
could we just … get rid of the boilerplate?
Slide 85
Slide 85 text
@holly_cummins
#RedHat
example: logging without quarkus
package com.example;
import org.jboss.logging.Logger;
public class MyService {
private static final Logger log = Logger.getLogger(MyService.class);
public void doSomething() {
log.info("It works!");
}
}
Slide 86
Slide 86 text
@holly_cummins
#RedHat
example: logging without quarkus
package com.example;
import org.jboss.logging.Logger;
public class MyService {
private static final Logger log = Logger.getLogger(MyService.class);
public void doSomething() {
log.info("It works!");
}
}
import io.quarkus.logging.Log;
Log
with quarkus
Slide 87
Slide 87 text
@holly_cummins #RedHat
takeaway
- think different
- go for radical waste reduction
Slide 88
Slide 88 text
@holly_cummins #RedHat
can you try too hard for efficiency?
Slide 89
Slide 89 text
@holly_cummins #RedHat
1700 1900 2000
Slide 90
Slide 90 text
@holly_cummins #RedHat
1700 1900 2000
Slide 91
Slide 91 text
@holly_cummins #RedHat
1700 1900 2000
Slide 92
Slide 92 text
@holly_cummins #RedHat
1700 1900 2000
Slide 93
Slide 93 text
@holly_cummins #RedHat
1700 1900 2000
Slide 94
Slide 94 text
@holly_cummins #RedHat
1700 1900 2000
Slide 95
Slide 95 text
@holly_cummins #RedHat
taylorism
Slide 96
Slide 96 text
@holly_cummins #RedHat
taylorism
Slide 97
Slide 97 text
@holly_cummins #RedHat
taylorism
≃
Slide 98
Slide 98 text
@holly_cummins #RedHat
taylorism
≃
Slide 99
Slide 99 text
@holly_cummins #RedHat
Slide 100
Slide 100 text
@holly_cummins #RedHat
Slide 101
Slide 101 text
@holly_cummins #RedHat
“data-driven”
Slide 102
Slide 102 text
@holly_cummins #RedHat
measuring the efficiency of people
Slide 103
Slide 103 text
@holly_cummins #RedHat
Slide 104
Slide 104 text
@holly_cummins #RedHat
Slide 105
Slide 105 text
@holly_cummins #RedHat
Slide 106
Slide 106 text
@holly_cummins #RedHat
it’s easy to measure wrong
Slide 107
Slide 107 text
@holly_cummins #RedHat
you get what you measure
are you measuring what you actually want?
Slide 108
Slide 108 text
@holly_cummins #RedHat
when efficiency isn’t efficient
trade-offs
Slide 109
Slide 109 text
@holly_cummins #RedHat
automation may not save as
much time as you think
Slide 110
Slide 110 text
@holly_cummins #RedHat
https://www.sify.com/ai-analytics/the-hilarious-and-horrifying-hallucinations-of-ai/
Slide 111
Slide 111 text
@holly_cummins #RedHat
“What is the world record for crossing
the English channel entirely on foot?”
https://www.sify.com/ai-analytics/the-hilarious-and-horrifying-hallucinations-of-ai/
Slide 112
Slide 112 text
@holly_cummins #RedHat
“What is the world record for crossing
the English channel entirely on foot?”
“The world record for crossing the
English Channel entirely on foot is
held by Christof Wandratsch of Germany,
who completed the crossing in 14 hours
and 51 minutes on August 14, 2020.”
https://www.sify.com/ai-analytics/the-hilarious-and-horrifying-hallucinations-of-ai/
Slide 113
Slide 113 text
@holly_cummins #RedHat
“What is the world record for crossing
the English channel entirely on foot?”
“The world record for crossing the
English Channel entirely on foot is
held by Christof Wandratsch of Germany,
who completed the crossing in 14 hours
and 51 minutes on August 14, 2020.”
https://www.sify.com/ai-analytics/the-hilarious-and-horrifying-hallucinations-of-ai/
bullshit as a service
Slide 114
Slide 114 text
@holly_cummins #RedHat
Slide 115
Slide 115 text
@holly_cummins #RedHat
“The submission accused firms of involvement
in scandals that either didn’t exist or that they
had nothing to do with. It referenced partners
being dismissed by firms that had never
employed them.”
Slide 116
Slide 116 text
@holly_cummins #RedHat
Slide 117
Slide 117 text
@holly_cummins #RedHat
“illusion of efficiency”
Slide 118
Slide 118 text
@holly_cummins #RedHat
can we afford efficiency?
Slide 119
Slide 119 text
No content
Slide 120
Slide 120 text
energy efficiency
of programming
languages
Slide 121
Slide 121 text
@holly_cummins #RedHat
“Rust is the hardest programming
language up to that time I’ve met.”
-Vorner
https://vorner.github.io/difficult.html
Slide 122
Slide 122 text
@holly_cummins #RedHat
language design trade-offs
Slide 123
Slide 123 text
@holly_cummins #RedHat
language design trade-offs
low-level control
Slide 124
Slide 124 text
@holly_cummins #RedHat
language design trade-offs
low-level control
higher-level
abstractions
Slide 125
Slide 125 text
@holly_cummins #RedHat
language design trade-offs
low-level control
higher-level
abstractions
“zero overhead”,
performant
Slide 126
Slide 126 text
@holly_cummins #RedHat
language design trade-offs
low-level control
higher-level
abstractions
“zero overhead”,
performant
developer
productivity
Slide 127
Slide 127 text
@holly_cummins #RedHat
language design trade-offs
low-level control
higher-level
abstractions
“zero overhead”,
performant
developer
productivity
raw performance may not
actually matter, but all
developers pay the cost of
the extra cognitive burden
Slide 128
Slide 128 text
@holly_cummins #RedHat
it is not sensible to rewrite all our business
apps in rust, no matter how efficient it is
Slide 129
Slide 129 text
@holly_cummins #RedHat
super-optimised code is less maintainable
Slide 130
Slide 130 text
@holly_cummins #RedHat
super-optimised code is less maintainable
software
efficiency
Slide 131
Slide 131 text
@holly_cummins #RedHat
super-optimised code is less maintainable
software
efficiency
developer
efficiency
Slide 132
Slide 132 text
@holly_cummins #RedHat
super-optimised code is less maintainable
software
efficiency
developer
efficiency
trade-off
Slide 134
Slide 134 text
@holly_cummins #RedHat
make sure you consider all requirements
Slide 135
Slide 135 text
@holly_cummins #RedHat
#RedHat
Slide 136
Slide 136 text
@holly_cummins #RedHat
highly efficient
(optimum number of legs)
#RedHat
Slide 137
Slide 137 text
@holly_cummins #RedHat
less efficient
(more legs than needed)
highly efficient
(optimum number of legs)
#RedHat
Slide 138
Slide 138 text
@holly_cummins #RedHat
less efficient
(more legs than needed)
highly efficient
(optimum number of legs)
resilient
#RedHat
Slide 139
Slide 139 text
@holly_cummins #RedHat
less efficient
(more legs than needed)
highly efficient
(optimum number of legs)
resilient
#RedHat
Slide 140
Slide 140 text
@holly_cummins #RedHat
less efficient
(more legs than needed)
highly efficient
(optimum number of legs)
no resiliency resilient
#RedHat
Slide 141
Slide 141 text
@holly_cummins #RedHat
less efficient
(more legs than needed)
highly efficient
(optimum number of legs)
resiliency is inefficient
no resiliency resilient
#RedHat
Slide 142
Slide 142 text
@holly_cummins #RedHat
it’s the same for people
Slide 143
Slide 143 text
@holly_cummins #RedHat
‘le blurring’ can be harmful to health
Slide 144
Slide 144 text
@holly_cummins #RedHat
le fun-at-work can help productivity
Slide 145
Slide 145 text
job satisfaction is
the no. 1 predictor
of organizational
performance
Slide 146
Slide 146 text
@holly_cummins #RedHat
research shows:
fun is good for business.
Slide 147
Slide 147 text
@holly_cummins #RedHat
research shows:
fun is good for business.
Slide 148
Slide 148 text
@holly_cummins #RedHat
studies show…
Slide 149
Slide 149 text
@holly_cummins #RedHat
studies show…
employee
fun
Slide 150
Slide 150 text
@holly_cummins #RedHat
studies show…
employee
fun
Slide 151
Slide 151 text
@holly_cummins #RedHat
studies show…
employee
fun
harder
working
Slide 152
Slide 152 text
@holly_cummins #RedHat
studies show…
employee
fun
harder
working
more
productivity
Slide 153
Slide 153 text
@holly_cummins #RedHat
studies show…
employee
fun
harder
working
more
productivity
less sick leave
Slide 154
Slide 154 text
@holly_cummins #RedHat
https:/
/hbr.org/2012/01/positive-intelligence
Slide 155
Slide 155 text
@holly_cummins #RedHat
“Your brain at positive is 31%
more productive than your
brain at negative, neutral or
stressed. "
https:/
/hbr.org/2012/01/positive-intelligence
Slide 156
Slide 156 text
@holly_cummins #RedHat
Slide 157
Slide 157 text
@holly_cummins #RedHat
https:/
/www2.warwick.ac.uk/fac/soc/economics/staff/eproto/workingpapers/happinessproductivity.pdf
Slide 158
Slide 158 text
@holly_cummins #RedHat
"Individuals [who just watched a
comedy video] have
approximately
12%
greater productivity."
https:/
/www2.warwick.ac.uk/fac/soc/economics/staff/eproto/workingpapers/happinessproductivity.pdf
Slide 159
Slide 159 text
@holly_cummins #RedHat
play helps creativity
we work in a creative industry
Slide 160
Slide 160 text
@holly_cummins #RedHat
doing nothing helps creativity
Slide 161
Slide 161 text
@holly_cummins #RedHat
the default mode network
idle minds can solve hard problems
Slide 162
Slide 162 text
@holly_cummins #RedHat
psychology says we need idle time; maths agrees
Slide 163
Slide 163 text
@holly_cummins #RedHat
queueing theory says systems need to run under-capacity to cope
psychology says we need idle time; maths agrees
Slide 164
Slide 164 text
@holly_cummins #RedHat
queuing theory basics
Slide 165
Slide 165 text
@holly_cummins #RedHat
queuing theory basics
arrival process
Slide 166
Slide 166 text
@holly_cummins #RedHat
queuing theory basics
arrival process
Poisson distribution
Slide 167
Slide 167 text
@holly_cummins #RedHat
queuing theory basics
queue
arrival process
Poisson distribution
Slide 168
Slide 168 text
@holly_cummins #RedHat
queuing theory basics
queue
arrival process
Poisson distribution
Slide 169
Slide 169 text
@holly_cummins #RedHat
queuing theory basics
queue
arrival process servers
Poisson distribution
Slide 170
Slide 170 text
@holly_cummins #RedHat
queuing theory basics
queue
arrival process servers
Poisson distribution
Slide 171
Slide 171 text
@holly_cummins #RedHat
queuing theory basics
queue
arrival process servers completed work
Poisson distribution
Slide 172
Slide 172 text
@holly_cummins #RedHat
queue
arrival process servers completed work
if server capacity is too low, wait times are high
Slide 173
Slide 173 text
@holly_cummins #RedHat
if arrival rates are low, servers will be idle
queue
arrival process servers completed work
Slide 174
Slide 174 text
@holly_cummins #RedHat
utilisation
lead time
http://brodzinski.com/2015/01/slack-time-value.html
assuming Poisson
distribution of arrivals
Slide 175
Slide 175 text
@holly_cummins #RedHat
utilisation
lead time
http://brodzinski.com/2015/01/slack-time-value.html
assuming Poisson
distribution of arrivals
80% utilisation → 90% utilisation:
wait times double
Slide 176
Slide 176 text
@holly_cummins #RedHat
utilisation
http://brodzinski.com/2015/01/slack-time-value.html
cost
delay
cost
assuming Poisson
distribution of arrivals
80% utilisation → 90% utilisation:
wait times double
Slide 177
Slide 177 text
@holly_cummins #RedHat
utilisation
http://brodzinski.com/2015/01/slack-time-value.html
cost
delay
cost
cost of idle
capacity
assuming Poisson
distribution of arrivals
80% utilisation → 90% utilisation:
wait times double
Slide 178
Slide 178 text
@holly_cummins #RedHat
trains could go faster than
the schedule needs them to
Slide 179
Slide 179 text
@holly_cummins #RedHat
speed
trains could go faster than
the schedule needs them to
Slide 180
Slide 180 text
@holly_cummins #RedHat
speed
vs
predictability
trains could go faster than
the schedule needs them to
but then any minor
delay is irrecoverable
Slide 181
Slide 181 text
@holly_cummins #RedHat
this is great
Slide 182
Slide 182 text
@holly_cummins #RedHat
this is great
relaxing has business value
Slide 183
Slide 183 text
@holly_cummins #RedHat
takeaway
- measure thoughtfully; you will get
what you measure
- remember the paradoxes:
- efficiency is not always efficient
- fun and idleness improves efficiency
Slide 184
Slide 184 text
@holly_cummins #RedHat
are you optimising the right thing?
efficiency can take unexpected forms
Slide 185
Slide 185 text
@holly_cummins #RedHat
what problem are
we trying to solve?
Slide 186
Slide 186 text
@holly_cummins #RedHat
what problem are
we trying to solve?
efficient
cool and bouncy
Slide 187
Slide 187 text
@holly_cummins #RedHat
what problem are
we trying to solve?
efficient
cool and bouncy
more efficient
squashy and blobby
Slide 188
Slide 188 text
@holly_cummins #RedHat
tl;dpa ⁃ our processes, development, and
software can all be more efficient
⁃ do look for waste, and get rid of it
⁃ don’t treat people like machines
⁃ beware the illusion of efficiency
⁃ remember that people and systems
function better with some idle time
(too long; didn’t pay attention)