2. Cosmologies
This presentation is about belief
systems – ‘cosmologies’ – and our
most basic assumptions about the This is the motion of Mars
nature of the (networking) universe. across the sky, as seen on
successive nights.
Do our models explain observed reality? If not,
the model has scope to be improved.
3. Reality vs Model
The ancient Greeks struggled to
explain this retrograde motion.
Indeed, the word ‘planet’ means
‘wanderer’.
4. Geocentric
Their basic assumption was that the
Earth is the centre of the universe.
This is an anthropocentric viewpoint.
5. Everything revolved around the
Earth, because that’s where we
are, and what could be more
important than us?
6. Geocentric
So the explanation for the motion
of planets was to appeal to
Gods! unknown and unknowable forces.
7. Then this guy comes along with a
better idea in the 2nd century AD.
Ptolemy
8. Geocentric
The planets go in little loops as
they go round the Earth.
Epicycles! All is explained!
9. Except that didn’t really account
for what we observed. There
was a residual motion that
wasn’t explained by epicycles. So
the basic assumptions had to be
re-visited.
Copernicus
A mere one and a bit millennia
later, Copernicus suggests that
maybe we (and the Earth) aren’t
the centre of the universe after
all.
10. Geocentric Heliocentric
His heresy was to say that the
Sun was the centre of the solar
system, and everything went
around that.
Which is a shift in the origin
of how we reasoned about
planetary motion, and
required us to let go of the
#fail assumption the whole
universe revolves around
us.
11. Isaac
Newton
This process of refinement
continued.
12. Geocentric Heliocentric
Gravity explained the
elliptical motion.
#fail Gravity
13. The Sun turned out not to
be the centre either – as
we understood galactic
structures.
14. And there were still
unexplained residual
differences between
observed reality and our
models. So relativity fills
that gap.
You Know Who
15. Phantom Dark Energy Model
(2007)
This refinement of
cosmological models
continues to this day.
16. Networking Cosmologies
I am proposing that we need to
shift our ‘origin’ and adopt a
better cosmology in order to
make progress in networking.
The existing A suggested
dominant improved
cosmology… cosmology…
“Obese” “Lean”
17. The rest of this
presentation
involves Agenda
comparing a lot
of ‘obese’ and
‘lean’
cosmologies.
“Lean Networking”
Implications
Lean WHAT WHY
Conversations & (“Obese
HOW Pipes”)
Lean Future of
Manufacturing Internet, Distributed
WebRTC, Network
policy Computing
19. This book introduced the
basic principles that we are
interested in, and was a
‘cosmological shift’ in
process manufacturing.
Theory of
Constraints
20. Bottleneck
Every production system
has stages of processing
raw inputs into valued There is always a
outputs. bottleneck somewhere
in the system.
Supply of Demand
raw for finished
materials goods
21. Different Worldviews
Batches make sure Cost Throughput
there’s no idle time.
Variation just drives
Accounting Accounting
up unit cost.
Focus on Utilisation (Cost) Flow (Throughput)
Work unit Batch Single-piece
Variation Bad Good
Key metric Cost per Unit Lead Time
Variation in quantity
and quality of
demand is ordinary.
22. Goldratt’s great Aha! Fallacy
insight was to
identify this fallacy.
A resource standing
idle is a major waste.
WRONG!
23. Journey to ‘Lean Anything’
HIGH But you can’t get flow
Resource efficiency
efficiency this way. Coping
with variation requires
slack, and you don’t have
Are the any.
machines
busy?
The natural tendency of
managers seeing idle men
and machinery is to make
LOW it get busy.
Is stuff LOW HIGH
produced fast Flow efficiency
enough?
24. Eliyahu M. Goldratt
“A plant in which everyone
is working all the time
is very inefficient.”
Busy machinery isn’t the
same as creating value,
which is from products
delivered to customers
on time.
25. Inventory
Keeping machinery busy
(WIP)
upstream of a bottleneck just
creates work-in-progress
(WIP), which ties up capital,
requires storage, and inhibits
flexibility.
26. Journey to ‘Lean Anything’
HIGH So to get to
Resource efficiency
production nirvana,
you need to master
flow efficiency (i.e.
low lead time) first,
then figure out
resource efficiency.
LOW
LOW HIGH
Flow efficiency
27. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT! Goldratt gave us a
methodology, which I
have grossly
oversimplified to these
two steps.
28. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT!
29. Cumulative Flow Diagram
There are flow
visualisation techniques
which emphasise the
metrics we care about:
lead time and WIP.
Source: agilemanagement.net
30. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT!
31. Regulate
Limit WIP by preventing
more inflow into the
Inflow
system than the
bottleneck can process.
32. Before we start on
networking, let’s
practise these
techniques. How could we use
lean principles to
escape from email
hell?
WARM-UP EXAMPLE
LEAN CONVERSATIONS
33. We’re not alone in
thinking about this
problem in this way.
“A lesson
from lean
manufacturing”
34. Aha! Fallacy
A human standing
idle is a major waste.
WRONG!
35. Bottleneck
You really don’t want to
spend your life doing
email. And even if you do,
there’s a finite amount of
time resource to dedicate
to the cause.
24
36. Journey to ‘Lean Conversations’
HIGH
Resource efficiency
So how can we lower
lead time and WIP to
create flow efficiency in
our email handling?
LOW
LOW HIGH
Flow efficiency
37. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT!
38. My inbox
I don’t keep my emails
staring at me from my
inbox, as if it was a to-do
list anyone can scribble
Empty… on.
(except new messages since I got off the plane)
39. I use a product to
visualise and prioritise
my work – and at a glace
see what needs to be
done next.
Kanban visualisation system
41. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT!
42. Limited WIP
My personal
workflow
I limit WIP to force
me to recognise my I also force myself
finite work capacity. to focus on a
maximum of two
activities at once.
My attention is the
bottleneck.
43. Regulate
Inflow
How can we do this
for email?
24
44. End ‘inboxbloat’ by limiting WIP
“Jane’s inbox has reached its limit of 30 unread items.
To resubmit your message at a later time, click here.”
“You have sent 3 messages to John.
He has a limit of 2 messages outstanding per person.
Which message do you want to delete?”
Holiday hell: “You have 319 new messages.”
– Just send them all back and ask people to re-submit!
45. That was the salad.
INTRODUCING Now for the burger.
LEAN NETWORKING
46. Three kinds of network
Network of Network of
Possibilities Promises
Network of
PROBABILITIES
TDM circuit world,
IMS world
NON-DETERMINISTIC STOCHASTIC DETERMINISTIC
Manage priority, Manage sessions,
control loops
Manage flow, admission
contention and capacity
Internet world
and trade loss/delay
47. UK demo ISP running since 2007
My colleagues have built
something that appears
impossible in the existing
cosmology.
Bounded flow isolation Unconditionally stable
Assured loss and delay Scavenger traffic
The benefits of connections and
connectionless… without needing circuits to
reserve capacity or congestion control loops!
48. Aha! Fallacy
How? Same as Goldratt –
they saw the fallacy
common to all
networking.
A network element standing
idle is a major waste.
WRONG!
49. Journey to ‘Lean Networking’
Stop making HIGH
Resource efficiency
networks
mindlessly busy
creating WIP
Flow efficiency in
networks means
containing loss and
delay to meet the
bounds required by the
application.
LOW
LOW HIGH
Flow efficiency
50. Key features of methodology
1. Visualise yourworkflow
2. Limit work-in-progress (WIP)
3. ERR… THAT’S IT!
54. Delay = WIP
Waiting packets are the
networking equivalent of
WIP.
55. WIP, WIP, WIP
Today’s networks
maximise WIP, with
TCP/IP being the worst
offender.
56. We use the word ‘idle’ to
mean ‘lazy’ as well as to
describe a network link
that is not transmitting.
This gives us a clue we
have made an
inappropriate moral value
judgement
57. Bertrand Russell
“Immense harm is
caused by the belief
that work is virtuous.”
What underpins this is a
belief that it is morally
wrong for links to be idle.
58. Limit WIP!
Instead we should build
networks so that they
limit packet ingress to the
known downstream
bottleneck, and minimise
WIP.
59. Work-
That means the most basic
assumption built into
every network router is
wrong – or at least
unhelpful and costly.
Conserving
Queues
Work-conserving queues are ones which
always transmit if the link is idle and a
packet is waiting.
60. Category Error
This is a big deal.
Networks don’t
The clue seemed to
do work!
be in the name –
‘net-work’ – but they
aren’t machines that
do work.
They do translocation.
61. ! Work metaphor !
Putting stuff in the wrong place stops
you doing something useful, i.e.
moving packets can create
negative value
A lot of the ‘work’
routers do in fact
destroys value.
65. Shannon solved this
‘noise’ stuff for single
links; we’re doing it What is ΔQ?
for complete
multiplexed systems.
An idea like
entropy or noise…
…with many mathematical
representations
66. Alan Turing
assumed Zen of Netwrking
this.
An ideal network copies data
instantly and perfectly
Real networks instead create
loss and delay
And that’s it.
All they do is
degrade.
67. Zen of Netwrking
Networks are
trading spaces
This is They aren’t pipes. They
important. just allocate the
Pay attention. degradation compared to
‘instant and perfect’.
68. Zen of Netwrking
Networks are
Which can be
traded (e.g. drop
trading spaces
more packets
means less delay
for the rest) that allocate
{loss-and-delay}
69. Zen of Netwrking
Networks are
trading spaces They allocate
this
that allocate
disappointment
(degradation) to
different flows.
disappointment
70. Zen of Netwrking
Networks are
trading spaces
that allocate
Which we give a
fancy name
quality attenuation
71. Zen of Netwrking
Networks are
trading spaces
that allocate
ΔQ And make it
harder to type.
72. The Three Fundamental
Laws of Networking
If you offer a load to a network,
1. ΔQ exists loss and delay will happen.
2. ΔQ is conserved You can’t un-lose or
un-delay stuff.
3. ΔQ has two degrees of freedom (for
mutable contention)
These are the ‘thermodynamic’ laws we
Rather like pressure-volume- have to work with and constrain the best
temperature you learnt in school outcome we can deliver.
for a piston, networks have load-
loss-delay.
73. So we are working with a different
mental model that shifts our origin.
Bandwidth ΔQ
Networks aren’t about pushing
packets along one behind another. Instead, networks just do inter-
That’s an anthropomorphic model process communications with
that wrongly treats packets like added loss and delay.
physical packages.
Inter-process
“Beads on a string”
comms
74. Bandwidth ΔQ
∞
In the bandwidth model, a packet is
0
Our origin is ‘zero loss and delay’. That
makes the mathematics tractable –
assumed to stay at rest until we do rather like putting the sun at the centre
‘work’ to push it along. The default is of elliptical orbits and forgetting
it takes an infinite time to get there. epicycles.
75. People talk about ‘quality of
service’, but what is this
‘quality’ they talk of?
Bandwidth ΔQ
Quality is the Quality is the
presence of absence of
something something
positive negative
Their quality is “priority over Our quality is “the absence of
other packets to get work excessive disappointment”.
done”.
76. Bandwidth ΔQ
“Megabytes per second” Loss/delay probability
distributions.
Averages Instants
77. Bandwidth ΔQ
Doing work that creates
negative value is dumb.
Stop doing it!
Work-conserving Non-work
queues conserving queues
79. Two kinds of obesity
The Internet folk want you to
Network of wildly overprovision networks Network of
Possibilities to make quality-demanding Promises
applications work. This is
expensive.
The traditional
telco folk want to
build lots and
lots of parallel
NON-DETERMINISTIC networks to DETERMINISTIC
separate out
Overprovision! different quality
requirements. Overbuild!
This is expensive.
(This doesn’t actually work – but they haven’t
figured it out yet, so let’s keep it a secret
between you and me.)
80. Two kinds of obesity
Network of
HIGH YOU
Possibilities
Resource efficiency
ARE
HERE
The Internet is stuck here. TCP
is designed to saturate
networks – i.e. create
resource efficiency. It can’t
create flow efficiency.
LOW
NON-DETERMINISTIC
Overprovision! LOW HIGH
Flow efficiency
81. We have lost control over the
customer experience
Network
PREDICTABLE CHAOTIC components
within-bounds quality
often behave
Transported load at
HIGH The Internet is poorly under
also unstable at high load
all loads and
suffers
congestion
Necessary to
collapse effects.
manage the
load on the
network to
remain within
LOW the predictable
region of
operation
LOW HIGH
Offered load
82. Two kinds of obesity
But they’ve gone totally
the wrong way since
Network of
HIGH
Promises
Resource efficiency
then with IMS which
creates a horribly
Traditional bloated and inflexible
telco folk were network.
heading the
right way with
technologies
like ATM.
YOU
LOW ARE
HERE DETERMINISTIC
LOW HIGH Overbuild!
Flow efficiency
83. Network obesity means
LIFESTYLE CHANGES!
Nobody wants to Aspiration
hear the lean Quantify desired outcomes e.g. how often is it OK
networking doctor for YouTube to show
tell them that their
morbid obesity
Expectation the “circle of death”?
Define quality need Average time to first
demands lifestyle frame? MOS score for
changes. VoIP apps?
Execution
Match delivery to requirement
How do aspirations map How to make
to statistical bounds on
loss and delay?
Assurance applications fail in the
Closing the loop right order when
instantaneous demand
Did it do what you exceeds supply?
expected? How do you
know?
84. This is a new
technology to put
real-time two-way
voice and video into
web browsers.
SO WHAT?
INTERNET, WEBRTC &
POLICY
85. Bufferbloat
The phenomenon
that Internet
This problem won’t
engineers struggle to
find a resolution
explain and resolve
without a change in
in their current
cosmology; attempts
cosmology is called
to fix it will spawn
‘bufferbloat’.
new problems.
86. We’re not measuring
The network the things that
Hierarchy of Need actually matter,
Applications need
finite bounds on loss Rapid variation in loss
and delay to work. and delay cause
application failure.
3. Uncontended: Low loss and delay
!
2. Stable: High stationarity
!
1. Feasible: High capacity
The bandwidth
school of thought
Note: exact requirements are application-dependent focuses entirely on
capacity. This is not
enough.
87. The Internet has an inbuilt …because TCP breaks basic
The Internet’s
‘statistical noise’ problem… control theory principles.
‘Global Warming’ problem
We’re multiplexing more
Dropping flow isolation apps, devices and users
together; and doing more
real-time applications as
+ Rising peaks of failure well as bulk data.
So bufferbloat-like effects
+ Loss of stationarity good.
Not cause more problems.
= More ‘extreme weather events’
(and poor for real-time comms)
The Internet is intrinsically unsuitable for
carrying all of society’s real-time
communications needs.
88. The guiding ‘end-to-end principle’ on
which Internet architecture is based is
a post-rationalisation of bad design
decisions from the 1970s.
This has a cost our kids will carry.
End-to-end
principle
89. Network Neutrality is
‘beads on a string’ model
Has broken philosophical foundations:
Value comes from flows with bounded ΔQ
– Not individual packets
Loss is OK and delay can be good!
– It’s not a fault or moral failure to avoid work
System of two degrees of freedom (loss/delay)
Network neutrality is – Not just one (delay)
intellectual nonsense. But
once most people thought it
was obvious the Sun went
round the Earth too.
90. ! Network Neutrality !
In man vs mathematics,
the mathematics takes a
really hard There are rigorous and
meaningful ways of defining
non-discrimination.
negotiating stance
Just not this way.
Network neutrality as currently conceived puts regulators in conflict
with the mathematics of statistical multiplexing. The maths will win.
92. Technological Revolutions &
Techno-economic
Financial Capital
revolutions go through Carlota Perez
this predictable cycle.
Electricity, Steel
& Heavy
Engineering
IT & Telecoms
Steam, Coal, Oil, Petrochemicals ? Biotech,
Iron, Railways & Automobiles Nanotech
1770 2012
93. The Turning Point
Networking becomes
You are here! c.70 years embedded into
after transistor everyday life, invisible
invented. in the way electricity is.
Amazing! Fit-for-purpose
It works!
We’re done inventing
the basic component
technologies –
datagrams,
smartphones, etc.
94. The Turning Point
ΔQ model fills the
intellectual hole Turing
Networking has been
left behind when he
done on a “try it and
assumed translocation
see” basis, without
was instant and perfect.
rigorous mathematical
foundations.
Networking Networking
as alchemy as chemistry
“Translocatability” now
has intellectual
foundations as solid as
“computability”.
95. Obese monoservice networks
Which means we can begin to
treat network obesity caused by
excessive consumption of
monoservice (single class)
networks.
96. End the Cycle of Failure
Select new Build
technology network
Tweak it Because that really isn’t Tweak network to
again working any more. get apps working
Build it Get
bigger complaints
97. What we need is…
Ubiquitous data
transport for a
digital society
And will be even less
suitable for what we need
in future.
98. Lean polyservice networks
We need to go from ‘black and
white’ to ‘colour’ networks. Each
flow has a different quality
(‘colour’) need.
99. Beyond cloud & internet:
Distributed Computing
ΔQ is only half of the puzzle. The
Internet is mis-architected in
other ways. Check out Recursive
ΔQ
RINA
Internet Architecture for how it
should be done.
100. Coming in 2013
New book by Martin Geddes:
“The Internet is
Just a Prototype”
101. Free newsletter
www.martingeddes.com
Congratulations, you have
survived this intellectual assault
course. For further intellectual
assaults, read my newsletter.
102. Future of Voice
& Telco-OTT workshops
London, England
October 23rd & 24th
www.futureofcomms.com
For a more gentle re-shaping of
your views on communications,
come to a public workshop.
103. Get in touch
Martin Geddes Pleased to hear from you.
Founder, Martin Geddes Consulting Ltd
Twitter: @martingeddes
Email: mail@martingeddes.com
Additional credits and thanks to:
Neil Davies, Peter Thompson, John Day, Fred Goldstein