85
Future of Broadband Workshop ITU Telecom World 2013, Bangkok Dr Neil Davies Predictable Network Solutions Ltd Martin Geddes Martin Geddes Consulting Ltd PREDICTABLE NETWORK SOLUTIONS © 2013 All Rights Reserved

Future of Broadband workshop presentation - ITU Telecom World 2013

Embed Size (px)

Citation preview

Future of Broadband Workshop ITU Telecom World 2013, Bangkok

Dr Neil Davies Predictable Network Solutions Ltd

Martin Geddes

Martin Geddes Consulting Ltd

PREDICTABLE

NETWORK

SOLUTIONS

© 2013 All Rights Reserved

Dr Neil Davies Co-founder and Chief Scientist Ex: University of Bristol (23 years).

Former technical head of joint university/research institute (SRF/PACT).

The only ex-ante network performance engineering company in the world.

Martin Geddes Founder Ex: BT, Telco 2.0, Sprint, Oracle, Oxford University.

Thought leader on the future of the telecommunications industry.

Consultancy on the future of voice, cloud and broadband.

PREDICTABLE NETWORK

SOLUTIONS

Our offer to you today

• Help you to understand the mismatch between:

– What people are aspiring to achieve (demand)

– What you are actually doing (supply)

• Propose how to close the gap

– Technically grounded in reality

– Practical advice on how to proceed

This may be a difficult message to hear

We’ve had a lot of experience of

people not wanting to hear

what we’re about to tell you

Our three key messages

1. Speed (‘bandwidth’) is no longer a helpful model for broadband.

2. The pursuit of ever more speed means the broadband business is in a death spiral.

3. You need to re-frame your resource model to survive.

ADDICTED TO PEAK SPEED

The whole market is

Selling data speed and mechanisms

A proxy for ‘more speed’

Pressure from regulatory environment for speed

SPEED!

League tables for speed

Incentives matter and

everyone gets incentivised to deliver more

speed

THING THAT MATTERS

Speed is not the only

Example: Satellite in Asia-Pacific

SERVICE A SERVICE B

< 1Mbit

> 6Mbit

Which is better?

Service A: Low variability

Service B: High variability

Same satellite, same location, similar time,

different service

More speed is not necessarily better

SERVICE A SERVICE B

‘Slower’

and good QoE

‘Faster’

but poor QoE

Probably not what you

would have expected

ALL BROADBAND DELIVERY

This is a common issue to

Was this just a one-off issue?

No.

DSL: same bandwidth, different QoE

The one on right has 1/3 the capability of the left for carrying POTS-quality VoIP

Comparison between two LLU broadband providers to same location in the UK

Two customers serviced off the same

pole in the same street by two

different wholesale DSL providers

Same problem on cable

We see the same thing on

other networks (e.g. 3G, small cells) but cannot share the data for contractual

reasons

GOOD USER EXPERIENCE?

What network attributes drive

A Skype experience (3-way call)

20M/2M Cable broadband 10M/1M ADSL Business LLU

1.8M/448k ADSL – wholesale 20CN

Loss: <0.5%. Delay (one-way): 50ms-60ms, jumping

to 500ms for a second or two, then back

Loss: Wandering from a typical 0%-2% up to as high as 48% for a second or two.

Delay: 50-70ms

Loss: 0.1%. Delay: 40ms-50ms

We measured path loss and

delay (summary on next slide)

Different speeds & characteristics

VERY FAST

& VARIABLE DELAY

FAST

& VARIABLE LOSS

SLOW & LOW VARIABILITY OF LOSS/DELAY

Good Experience

Bad Experience Bad Experience

Speed was not the key differentiator

Bad Skype QoE

SPEED

VA

RIA

BIL

IT

Y

SLOW FAST

LOW

HIGH

The faster broadband lines

gave a worse experience as

reported by Skype’s own QoE metric

Why did these user experiences differ?

Because they had different

loss and delay (and that’s it!)

So why are we promoting

‘speed’?

The application hierarchy of need

3. Predictable loss and delay

2. Stable: Good ‘stationarity’

1. Feasible: Sufficient capacity

Note: exact requirements are application-dependent

!

!

These are not getting enough

attention

We need more than just

‘speed’ for good QoE

Yes, we need capacity

DRIVE COST?

What network attributes

Valid reasons for spending capex

• More customers

More revenue

• Increased usage

More revenue

• Regulatory requirement

Invalid reason for spending capex

Premature infrastructure and capacity upgrades

Lots of capex spent on ‘tin’

Spectrum, fibre, copper, ducts,

street cabinets, cell towers, and the transmission

and routing equipment

Is it being well spent?

More, more, more

More supply

More elastic demand

Faster saturation of

backhaul

More variability

Lower QoE

More complaints and churn

When we believe more speed is the only answer, we

are doomed to go round again (and

again)

There is a ‘jackhammer’

effect that gets worse over

time

Failure of technology to keep

up with ever rising demand

forces shorter upgrade cycles

Rising load makes

service quality fall,

forcing upgrades

Serv

ice

Qu

alit

y

Time

Un

dep

reci

ated

Ass

et V

alu

e

Time

The investment ‘cycle of doom’

Death via unserviceable

debt load

QoE declines faster than

network planning rules forecast Se

rvic

e q

ual

ity

Un

dep

reci

ated

as

set

valu

e

Telecoms is a capital killer

Source: PwC http://www.pwc.com/en_GX/gx/communications/publications/assets/pwc_capex_final_21may12.pdf

As an industry, we’re not

covering our cost of capital

Something is very badly wrong

PREMATURE UPGRADES?

What drives the

Qu

alit

y o

f ex

pe

rie

nce

LOW HIGH

LOW

HIGH

HEAVEN

How to think about cost drivers

HELL

Resource efficiency

Network has lots of users, who feel like the network is still empty

because they are suitably isolated from

each other

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW HIGH

LOW

HIGH

Access network

As load increases, QoE falls

Add more demand to today’s packet networks, and

everyone’s experience degrades since all your packets are ‘pollution’

to other users

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW MEDIUM

LOW

HIGH

HIGH

HEAVEN HEAVEN Heaven gets further away

Add capacity to resolve falling QoE

Access network

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW MEDIUM

LOW

HIGH

HIGH

Current approaches require all traffic to be schedulable within very short timescales

Can’t ensure QoE for applications with strong stationarity requirements

Access network

Infeasible

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW MEDIUM

LOW

HIGH

HIGH

The only way current network elements

can deliver good enough QoE is by

being idle frequently

Broadband networks need to be kept empty to keep working

! !

This microscopic queueing effect has

massive macroscopic implications, both

technically and economically

Queues need time to ‘relax’

Access network

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW MEDIUM

LOW

HIGH

HIGH

Protocols ‘collapse’ ‘Goodput’ plummets

Because of this you currently can’t run networks ‘hot’

Access network

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW MEDIUM

LOW

HIGH

HIGH

With every upgrade the QoE boundary for next upgrade drops

1 2 3 4 5

Diminishing returns from adding more

capacity to solve excess delay

UPGRADE THRESHOLD

RESOURCE PROBLEM?

How to re-frame the

Delay (and loss) have a structure

Delay is due to…

V S G Geography Serialisation

speed Variability

Why trust in increasing speed is now misplaced

G

Pre-IP Early IP Now

The speed of light is not changing

Geography

Why trust in increasing speed is now misplaced

Historically speed did correlate with

more value

G

S

Pre-IP Early IP Now

Geography

Serialisation speed

Why trust in increasing speed is now misplaced

G

S

V Variability

Pre-IP Early IP Now

Now dominates application

performance

Serialisation speed

Geography

Not all packets experience this

much delay, but the outliers are the

ones that matter to QoE

The commercial challenge:

How to break the investment cycle of doom?

The technical challenge:

How to measure, manipulate and

manage ‘V’?

COSTS

REVENUE

DATA FLOWS

APPLICATION OUTCOMES

FIT-FOR-PURPOSE EXPERIENCE

POWERED MECHANISMS

TRANSMISSION RESOUCE

REQUIRES

ENABLES

UNPOWERED TIN

SCHEDULING

Supply

-

Demand +

We need a robust causative model of

the relationship between

operator revenue and cost

Click here for separate presentation on this

model

Networks are ‘trading spaces’

How ‘V’ is distributed among competing streams

is how demand is matched to the supply

COSTS

REVENUE

SCHEDULING

Scheduling

This is where supply and demand meet

This makes all the difference between commercial success

and failure

…and nowhere else

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW HIGH

LOW

HIGH

IDEAL SCHEDULING

The real difference between telecoms heaven and hell

TODAY!

Your problem: magical thinking

When there is excessive delay, you are

trying to make V disappear by building more capacity rather than distributing it

through scheduling

Capacity demand LOW HIGH

Feasible Infeasible

MAX CAPACITY

TWO fundamental resource limits

Sch

ed

ula

bili

ty

de

man

d

LOW

HIGH If you want to move 10mbits in 1 sec, you need

(at least) 10mbit/sec of transmission

Capacity demand LOW HIGH

TWO fundamental resource limits

Feasible

Sch

ed

ula

bili

ty

de

man

d

Infeasible

LOW

HIGH

Even with perfect knowledge and

mechanisms, you can only schedule

so well

MAX SCHEDULABILITY

Capacity demand LOW HIGH

TWO fundamental resource limits

Feasible

Sch

ed

ula

bili

ty

de

man

d Infeasible

LOW

HIGH

In practise we aren’t nearly

that good

Feasible

Infeasible

Capacity demand LOW HIGH

Feasible Infeasible

MAX CAPACITY

TWO fundamental resource limits

MAX SCHEDULABILITY

Sch

ed

ula

bili

ty

de

man

d

LOW

HIGH

We typically hit this limit first (which is

why adding capacity is not a good idea)

Our problem

Schedulability demand is growing fast

VoIP, gaming, 2-way video, UC, HTML5 web…

The problem

Solving schedulability issues (i.e. non-stationarity)

with capacity is inefficient and ineffective

Resource efficiency

Qu

alit

y o

f ex

pe

rie

nce

LOW HIGH

LOW

HIGH

HEAVEN

There is only one feasible route

HELL If you

focus on resource

usage first…

Focus on scheduling and QoE

first

No slack means this is

not possible

…then you get stuck here

We are going about

broadband the wrong way

TECHNICALLY ACHIEVABLE?

Is the path to heaven

Yes

Pro-active control over scheduling ‘V’

We used a different resource model to achieve this

HELL

HEAVEN

90% of load

We built a demo ISP to

prove what we say actually

works This network is still delivering good QoE at 100% load

Some quality-insensitive traffic

gets slightly worse treatment

RESOURCE MODEL?

What is the right

Different supply ‘performance’

1.LOSS

2.DELAY

} Quality

Networks create value by moving data with

bounded loss and delay

This is the resource

model you need

Quality

Need to frame the supply differently to make issues soluble

Bandwidth

‘Bandwidth’ is the presence of

something wanted

‘Quality’ is the absence of something unwanted

Loss and delay

Speed

RESOURCE MODEL?

How to use this quality-centric

What has to change?

NOW FUTURE

SUPPLY-PUSH

Selling commodity bandwidth

inputs

DEMAND-PULL

Selling differentiated

application outcomes

What has to change?

Focus on enabling outcomes not higher ‘speed’

Properly characterise your demand

Demand-driven model

COSTS

REVENUE

FLOWS

OUTCOMES

FIT-FOR-PURPOSE EXPERIENCE

MECHANISMS

TRANSMISSION

REQUIRES

ENABLES

TIN

SCHEDULING

Construct Matching Supply

Characterise Demand

Do this first…

…then do this

What has to change?

Understand how delivered QoE is a function

of loss and delay

Properly characterise your supply requirements

So your service is fit-for-purpose

SUPPLY RESOURCE MODEL

A practical network

Example of a supply approach: Three layer model

Superior traffic costs more to deliver… so should attract a premium

Economy

Standard

Superior

Standard traffic is today’s off-peak Internet… but is consistently the same

Economy traffic does not drive capacity upgrades Today’s QoS

mechanisms don’t deliver

this (or create a service of no value trying)

They don’t understand the ‘trades’

properly

Extend this to a five class model

• Provisioning capacity for total resilience is a real cost…

– …but not everyone needs it

– So we separate out resilient traffic from non-resilient

• When there is a period of network stress, some people may get reduced service…

– …but they still have connectivity

Economy

Sup

erio

r St

and

ard

Sup

erior

Stand

ard

Res

ilien

t

Five class model has rational economics

Sup

erio

r St

and

ard

Su

perio

r Stan

dard

Economy

Sup

erio

r St

and

ard

Sup

erior

Stand

ard

Economy

Sup

erio

r St

and

ard

Sup

erior

Stand

ard

Economy

Drives capacity planning (primary service)

cost

Drives resilience & redundancy capacity

planning cost

Drives revenue

SUMMARY & CONCLUSIONS

We are in a race to the bottom

We’ve got into a fight with the

mathematics of statistical

multiplexing

This is not negotiable

Why so? Demand is not being met

Supply-push business and technology model

Why so? Wrong kind of supply

Failure to align with underlying and unchanging reality of packet

networking

Speed (and volume) are not value

Dangerous myth: MORE SPEED IS ALWAYS BETTER

We’ve seen networks

where adding capacity made performance

get worse

! !

Broadband is becoming critical national infrastructure

Needs to be dependable No ice cream

(or insulin) without fridge-freezers, which need a reliable power supply

Advanced services need

predictable and dependable

supply

We are creating a digital society

Implicit social contract

We can’t externalise

our collective risks

Keep getting scheduling wrong: Crisis of legitimacy

Angry: Customers Investors

Regulators Governments

Get scheduling right: Golden age of broadband

1890s

Railways

1920s

Electricity

1960s

Oil

2020s

Broadband

RECOMMENDATIONS

What operators should be asking themselves

1. Why am I trying to solve my scheduling problems with more capacity?

2. For my key customer applications, am I delivering the network supply that enables good QoE?

– i.e. am I delivering the right loss and delay?

3. Given that there is a trading space, am I constructing and offering the right data transport products?

What regulators should be asking themselves

1. What is the value that I am getting from demanding more speed?

2. Measurement is de facto regulation, therefore am I measuring the right thing?

3. What are the key applications that need managed QoE and cost to drive societal benefits?

To learn more

Free Future of Communications newsletter: www.martingeddes.com

Follow Martin Geddes on Twitter: @martingeddes

Other presentations: slideshare.net/mgeddes/

White papers on network performance: www.pnsol.com

Neil Davies [email protected]

Martin Geddes [email protected]

PREDICTABLE

NETWORK

SOLUTIONS