A cooperative approach to content delivery

A cooperative approach

to content delivery

We would like to thank Robert Kenny of Communications

Chambers for providing helpful comments on this briefing paper.

Cover illustration by Bratislav Milenkovic.

A cooperative approach

to content delivery

A Netflix briefing paper

2021

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Online work, school, and entertainment have

long been complements to their physical

counterparts. But with no other option, basic

needs were met for many of us with these not

always perfect replacements. The pandemic

highlighted how robust online alternatives can

be, as many quickly scaled to meet the demand

for interpersonal connection. Internet networks

had to adapt to meet the growing demand for

high speed broadband, by leveraging their

existing capacity for a massively increased load.

The increased demands during the pandemic

highlight the importance of cooperation.

Together, content providers and networks have

been meeting people’s needs. Without both, our

ability to sustain human connection, distanced

productivity, or continuous learning is challenged.

Both must be fostered in a sustainable way.

We saw examples of this cooperation in its

finest form during the pandemic. Where the

exceptional nature of the pandemic made it

Foreword

When the COVID-19 pandemic forced us to recede into a

global quarantine, the ways in which many of us connect with

each other - through work, school, and entertainment - ceased

in their typical forms. Where possible, we moved the essential

functions of society online using our internet connections.

necessary, Netflix and other video streaming

services reduced the bandwidth of streams to

decrease traffic volume, and ISPs around the

world quickly grew capacity to ensure continued

connectivity for internet users. To get content

to consumers more efficiently, Netflix and ISPs

worked together to deploy additional servers into

ISPs’ networks as well as increase capacity in our

backbone network and at local internet exchange

sites in order to fulfill the growing demands for

Netflix content.

An internet that can meet the needs of

a globally connected population - during a

pandemic and beyond - depends on such

cooperation between ISPs and content providers

to best provide the services they offer to

their customers.

Humanity will move beyond the pandemic.

But we should take the lessons we have learned

about the need for a collaborative approach to

online content delivery forward long into the future.

Gina Haspilaire

Vice-President, Content Delivery

Netflix

27Threats to the ecosystem:

Understanding traffic charges

1 2

15Executive summary

Content & network:

The need for cooperation

Conclusion

Glossary of terms

Appendix:

Transit and peering

Cooperation in practice:

Netflix Open Connect

Contents

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Consumers buy internet access from ISPs in order to reach

content and applications. Without content, consumers

would have less need for internet access. Likewise, without a

connected population, content providers would have no ability

to service consumer requests. Put simply, online services and

internet networks depend upon each other.

Executive

summary

This complementary relationship between

content providers and networks has resulted in a

model of content delivery based on cooperation

- a model which has promoted a virtuous circle

of innovation, whereby the improvements in

networks inspire new forms of content, and this

in turn drives uptake of higher speed broadband

plans. This model has also encouraged greater

connectivity and the wider societal benefits that

occur when consumers and businesses have

improved access to information and services.

However, despite its strengths, there have been

threats to this model. In some countries, proposals

have been made that could place a tollbooth in

front of the entrance to networks, blocking traffic

requested by consumers unless content providers

agree to pay to lift the gate. These proposals have

often been a response to the myth that content

and application providers do not contribute to the

costs associated with building out connectivity,

leaving ISPs unable to upgrade and expand

broadband networks.

Against this backdrop, this report demonstrates

that Netflix and other content providers are

investing significantly in content delivery

infrastructure and video streaming technology

to deliver content more efficiently. Further, it

shows that an approach built on cooperation,

rather than traffic charges imposed on content

providers by ISPs, is good for ISPs, content

providers, consumers, and society as a whole.

Chapter one of the report explores the

complementary relationship between ISPs

and content providers and shows how this

relationship generates significant benefits

for both. For content providers, largely

frictionless connectivity to a huge user base

A Netflix briefing paper

2021

A cooperative approach

to content delivery

has underpinned the uptake and usage of those

services. For networks, consumers purchase

internet access to reach such services.

Online entertainment services like Netflix

invest significantly in content to engage and

delight their customers. Netflix alone spent more

than $12.5B on content in 2020. Reaching

such content is a substantial portion of what

consumers do with their internet connection and

may prompt them to upgrade their broadband

connection, generating incremental revenue

for ISPs. Empirical evidence suggests that

broadband customers who use Netflix are more

likely to be on higher speed (and more expensive)

broadband connections. For example, a study

of broadband users in Italy in 2017 found that of

those on ADSL (Asymmetric Digital Subscriber

Line) connections, 12% were Netflix users, while

for those with FTTH (Fiber To The Home), 18%

were. In the UK, 77% of those with standard

broadband were users of subscription video-on-

demand services (such as Netflix), compared

to 86% and 90% respectively of those with

superfast broadband and ultrafast broadband.

In markets where broadband adoption has

not reached saturation, online services can

encourage broadband deployment and uptake.

In addition, in markets where fixed broadband

adoption is high, video traffic is a key reason for

customers to retain a fixed internet connection

rather than going mobile-only.

This complementary relationship incentivizes

ISPs and content providers to work together

to focus on the best technical approach for

consumers and this, in turn, generates value

for ISPs, content providers and consumers.

Chapter two of the report explains how Netflix

efficiently delivers its content. Netflix’s Open

Connect Content Delivery Network (CDN) moves

content closer to consumers using a distributed

network of local servers at the edge of, or within,

an ISP’s network.

This benefits ISPs by reducing the distance

they must go to fetch traffic, and thus their costs.

For our ISP partners that have opted to deploy

servers within their own networks, this cost

saving is estimated to be $1.2B in 2020.

Moving content closer also frees up long

haul network infrastructure for the other types

of traffic for which those links are essential - for

example, live video streaming or voice calls over

the internet which cannot be stored locally.

This can reduce congestion, resulting in a

higher quality service for consumers.

Netflix also invests in improving encoding

and video compression to deliver high quality

content without using unnecessary bandwidth.

As a result, the number of hours a Netflix

member can stream per GB of data has increased

by more than 200%

over the last five years.

Ofcom, Technology Tracker 2021, 17 December 2020.

Standard broadband is defined as providing a download

speed of less than 30 Mbps, superfast broadband between

30 Mbps and 300 Mbps and ultrafast broadband more than

300 Mbps.

Based on the latest encoding profile.

$1.2B

200%

on estimated savings for our ISP

partners that have opted to deploy

servers within their own networks

increase in the number of

hours a Netflix member can

stream over the last five years

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Chapter three concludes with an assessment

of why traffic charges imposed on content

providers by ISPs will be harmful to consumers.

It considers the key arguments cited by those

who support traffic charges: that ISPs need funds

from content providers for investment and that

content providers do not contribute to the costs

of delivering content.

It also discusses the main risks associated

with traffic charges.

Traffic charges pose problems because

there are no alternative routes to the end user.

A consumer’s chosen ISP has complete control

over what content can travel over the connection.

This creates a ‘terminating access monopoly’

which can pose multiple risks when combined

with traffic charges.

Traffic charges can lead to double-charging

where ISPs seek to charge content providers for

the carriage of traffic that their customers have

already paid for. Customers pay for an internet

connection on the understanding that it will allow

them to reach all content available on the entire

internet, not just from those content providers

who have agreed to their ISP’s traffic fees. The

problem of double charging is exacerbated when

the speeds required for the content being delivered

are well under the speeds paid for by consumers.

Traffic charges may also allow ISPs to leverage

their own video offers. Many ISPs compete

directly with streaming video services, either

because they are themselves a cable TV operator,

or a telco with an IPTV offering. Any degradation

of quality for other video providers strengthens the

competitive position of an ISP’s own TV service,

disturbing a level playing field.

Traffic charges can lead to perverse

incentives. The only way for an ISP to force a

content provider to pay for traffic charges is

to ensure congestion otherwise limits an ISP’s

customers from receiving requested traffic from

the content provider. This congestion represents

powerful leverage for the ISP to force the content

provider to accept the demanded cost to connect

directly, and thus the ISP has little incentive to

remedy the congestion.

Furthermore, allowing payments may

discourage content providers from investing

in moving content closer to consumers,

to the detriment of ISPs and consumers alike.

There is no requirement to move content closer

to consumers. An uncongested long haul

connection can still deliver high quality video.

If an ISP imposes charges, the content provider

may choose to deliver their traffic from

a jurisdiction where payments are not required.

For the ISP, this will likely mean that it is now

paying to receive the traffic, since it may need

to invest in international capacity to carry it back

to their consumers from that jurisdiction.

Consumers invariably lose in this situation.

They will be exposed to the risk of congestion

along the path to more remote content, with

possible degradation of their user experience.

There is also a risk that consumers will lose

access to the plurality of voices and applications

that have enabled the internet to thrive.

Thus traffic charges are a backward step,

with the potential to disrupt a highly productive

partnership between ISPs and content providers

that has been the bedrock of the internet’s

success to date.

Looking ahead, a cooperative model will only

be sustained under two conditions. Firstly if

content providers and ISPs continue to recognize

the complementary nature of networks and

content. Secondly, if governments acknowledge

the important role that content providers play in

stimulating demand for broadband services and

establish supporting policy frameworks which

restrict the imposition of charges by ISPs on

content providers.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Content

& network

The need

for cooperation

The complementary nature

of content and networks

Investments in video create

demand for connectivity

Efficiency benefits

of a cooperative model

A Netflix briefing paper

2021

A cooperative approach

to content delivery

The complementary nature

of content and networks

The very essence of the internet is that it is a

‘network of networks’ - not a monolithic single

network, but many separate interconnected

networks of completely different types.

The internet, for the most part,

has allowed

anyone online to communicate with anyone else

online. This flexibility has been transformative.

A consumer who purchases internet access

can reach nearly any type of content from nearly

any public network, anywhere around the world.

Consumers can access a plethora of information

and services and ultimately choose which

businesses are successful. As of 2020, there

were 4.7B people online.

There are 192M

active websites globally,

and 3M apps available

in the Google Play store.

For content providers, largely frictionless

connectivity to a huge user base has created

incentives for investment in applications, and

underpinned the uptake and usage of those

applications.

This in turn has led to an explosion

of innovation and competition.

For networks, consumers purchase internet

access to reach such content and applications.

And as those applications become more advanced,

demand for bandwidth has grown. The networks

have found great value in providing access to

information, entertainment, and each other.

The relationship between content providers

and networks is complementary and the

complementary relationship between ISPs and

providers of video content is particularly strong.

Services like Netflix rely on networks to distribute

content to consumers and ISPs benefit from

increased demand for connectivity.

For example, some governments block access to certain

sites in other countries, and many ISPs block illegal content.

Statistica, Worldwide Population as of October 2020,

Statistica 2021.

Netcraft, September 2020 Web Server Survey,

23 September 2020.

AppBrain, Number of Android apps on Google Play,

14 October 2020.

See, for instance, Plum, The open internet – a platform for

growth, 2011; TRPC, Fostering an Open Internet in Asia,

January 2017.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

12.5

$ B

spent on content in

2020, driving demand

for high speed internet

connections

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Streaming entertainment - which is often

on-demand, personalized, and available on

any screen - is expanding rapidly. Consumers

appreciate the freedom, flexibility, and rapid

innovation streaming entertainment services

afford. Access to high speed and reliable internet

and internet connected devices have enabled

these streaming services to become popular.

Watching video is now a substantial portion

of what many consumers do with their high

speed broadband connection. Video streaming

represents 58% of traffic. (YouTube is 16

percentage points of this, and Netflix 11 points).

This high level of usage is only brought about

because of substantial investments in content

that consumers want to watch.

Investment in video content

Netflix is one of the world’s leading entertainment

services. The strength of our global content offer

is fundamental to our success. Great content

grows engagement among our members, which

we believe drives word-of-mouth, improves

retention and grows memberships. In 2020,

we spent over $12.5B

on streamed content.

We invested in both Netflix originals and

second-run movies and TV shows across

many different genres (scripted series, films,

documentaries, comedy, unscripted TV, kids

and family, anime, etc). These movies and TV

shows come from producers all around the world,

and their stories are shared with the world.

Historically, television has been bound by a

linear program schedule and often a reliance on

advertising, which put a premium on airing content

that will appeal to the widest audience during

times likely to attract the most viewers. But online

business models, such as Netflix’s, depend on

consumers consistently finding exciting content

to watch, not just at prime time, but whenever the

consumer wants. This gives streaming service

customers the opportunity to discover new

content and enables streaming services to make

significant investments in innovative programming

from creators that historically may have had

difficulty reaching a global audience.

Netflix is not the only company investing in

video content. Unlike traditional broadcasters,

which were reliant on exclusive access to

spectrum or dedicated cables, streaming

services are able to offer their service via an

internet connection. These lower barriers to

entry for content delivery mean that diverse and

competitive players can offer innovative and

compelling content. We, like other major providers

of online entertainment services, are all operating

in a dynamic and thriving sector where a wide

range of players are incentivized to continually

innovate and invest in their services and the

audiovisual ecosystem in order to win consumer

attention. Many other major entertainment

Investments in video create

demand for connectivity

Sandvine, The Global Internet Phenomena Report, May

2020. Note that these figures relate to a period when many

were at home due to COVID-19. However, figures for total

streaming and for Netflix were broadly similar in 2019,

though YouTube was lower (9%).

See Netflix 2020 Quarterly Earnings Fourth Quarter

Earnings, Financial Statements. Cash spending on content

can be derived from the cash flow statement. The sum of

Additions to Streaming Content Assets and the Change in

Streaming Content Liabilities equates to cash spending on

streaming content.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

companies like Disney and NBCUniversal are

investing in their own streaming services, as

are Amazon and Apple, in order to compete

with providers of user generated content, video

games, and other online services for a share of

consumers’ time and spending.

Creating demand

for faster broadband

Video - and in particular higher definition

video - prompts consumers to upgrade their

broadband connection, generating incremental

revenue for ISPs. The FTTH Council, for

instance, has seen video growth as a driver of

fibre to the home (FTTH) adoption.

Ofcom,

the UK’s communications regulator, has found

that:

The main drivers of residential demand

for higher speeds have been an increase in the

use of video-on-demand and gaming, and the

simultaneous use in a home of multiple devices.

A report for BEREC (the association of

European regulators) made a direct link to

revenue:

Content developers and providers

of Over The Top (OTT) services such as the

various third-party video streaming services ...

are important [in] driving increased demand for

bandwidth but are not typically directly involved

in the delivery of that bandwidth, but ... could

certainly influence the operators’ ARPU.

And operators quite explicitly use video

streaming to persuade customers to take

higher speed plans, as the examples on

the next page show.

There is empirical evidence that broadband

customers who use Netflix, for example, are more

likely to be on higher speed (more expensive)

connections. A study of broadband users in

Italy in 2017 found that of those on ADSL

connections, 12% were Netflix users, while for

those with FTTH, 18% were

. (Note that there

was no material difference between ADSL and

FTTH usage rates for YouTube, suggesting

that it is not simply streaming video that drives

purchase of higher speed connections, but rather

HD and UHD video like that provided by Netflix).

In the UK, 77% of those with standard broadband

were users of subscription video-on-demand

services (such as Netflix), compared to 86%

and 90% respectively of those with superfast

and ultrafast broadband.

Thus video on demand, in common with other

forms of entertainment (such as gaming and

streaming live sports) is a key driver of demand

for higher speed broadband.

Encouraging broadband uptake

and broadband deployment

In markets where broadband adoption has not

reached saturation, video streaming services

can help to encourage uptake, with benefits

to domestic users (who can now interact

with the new users) and to businesses (who

can implement new processes). According

to the Commonwealth Telecommunications

Organisation:

OTT applications stimulate

broadband adoption and thus economic

growth and tax receipts.

IDATE for FTTH, FTTH Forecast for EUROPE, March 2019.

Ofcom, Promoting competition and investment in fibre

networks: Wholesale Fixed Telecoms Market Review 2021-

26. Volume 2: Market assessment, 8 January 2020

Decision Analysis, SPC Network, Strategy Dynamics &

Greenwood Strategic Advisors for BEREC, Study on the

determinants of investment in VHCN – a System Dynamics

approach. Volume 1: Technical Report, November 2019

Those using Netflix at least once per week. Trevisan, Martino

et al., Five years at the edge: watching internet from the ISP

network, April 2020.

Ofcom, Technology Tracker 2021, 17 December 2020.

Survey participants were asked which of these fixed

broadband services does your household have: standard

broadband (download speed of less than 30 Mbps);

superfast broadband (download speed is 30 Mbps or

higher and less than 300 Mbps); ultrafast broad broadband

(download speed is 300 Mbps or higher).

CTO, Over The Top (OTT) Applications & the Internet Value

Chain, 22 May 2020.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

ISPs using streaming as a prompt to upgrade to higher speeds

Mbps

100

Mbps

Standard Plus Evening Speed

Superfast Unlimited

Maxis Fibre

RECOMMENDED FOR:

• 5-7 people streaming in HD

• Responsive online gaming

RECOMMENDED FOR:

• Ideal for homes where

streaming the latest Netflix

series is the perfect night in

RECOMMENDED FOR:

• Streaming on multiple devices

• Sharing videos and photos

• Online gaming

RECOMMENDED FOR:

• Full HD streaming & browsing

• 2-3 users on up to 5 devices

• Double storey or condominium

Mbps

Maxis Fibre

Standard Fibre Unlimited

Standard Evening Speed

RECOMMENDED FOR:

• Light browsing & streaming video

• Single user on up to 2 devices

• Single storey or condominium

RECOMMENDED FOR:

• For doing all that you love

online without any data worries

RECOMMENDED FOR:

• HD video streaming

• Sharing videos and photos

• Online gaming

RECOMMENDED FOR:

• 1-3 people browsing

• Streaming in SD

• Online gaming

ISP websites [accessed 20 October 2020]. This is a brief summary of broadband

packages offered by third parties and not a detailed breakdown.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

The ITU has reported that:

Demand for OTTs

results in both new subscribers for broadband

services and existing subscribers upgrading their

subscriptions for greater speed and bandwidth.

In addition, in markets where fixed broadband

adoption is high, video traffic is a key reason for

customers to retain a fixed internet connection

rather than going mobile-only.

Increased uptake of broadband, and higher

speed broadband in particular, is clearly good for

ISPs but it also has spillover benefits to society

as a whole. Many governments have invested

substantially to improve the availability of

broadband, believing in its economic and social

benefits. But by itself, availability has no value - it

is consumer uptake that matters. If online content

providers drive uptake of beneficial broadband,

then that helps realise the value of governments’

investment in availability.

This complementary relationship incentivizes

ISPs and content providers to work together

to focus on the best technical approach for

consumers and this, in turn, generates value

for ISPs, content providers and consumers.

In addition, capabilities of the network inspire

new forms of content, which in turn encourage

uptake of faster and more robust broadband.

Video, for example, is a key reason for users to

upgrade their fixed broadband speed or take

fixed broadband in the first place.

In the vast majority of cases, this virtuous

circle results in a cooperative model of

deployment where content providers and

networks work hand-in-hand to deliver content

more efficiently, reducing costs for both parties.

As we discuss in the next section, Netflix invests

substantially to deliver its traffic in such a way as

to make it easier for ISPs.

Efficiency benefits

of a cooperative model

This cooperative model works best

when neither side seeks to charge the other.

For instance, Netflix (like many other online

content providers) does not try to charge ISPs.

Similarly, the vast majority of ISPs around the

world do not charge content providers for delivery

of their traffic, though as discussed in Chapter 3,

a few wish to change this.

Today’s approach is the result of decades of

thoughtful development by content providers

and ISPs. It is also an approach that ultimately

provides benefits for consumers.

We now turn to a description of how Netflix

(and others) deliver their content.

International Telecommunication Union, Economic

impact of OTTs on national telecommunication/ICT

markets, 15 June 2020.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Cooperation

in practice

Netflix

Open Connect

Moving content closer

Description of Netflix Open Connect

Benefits for ISPs from Open Connect

Encoding

A Netflix briefing paper

2021

A cooperative approach

to content delivery

In the early days of the internet, content would

typically sit on a single server, and then travel

over multiple networks to reach consumers. ISPs

either had to build out significant infrastructure

to reach the content, or else they relied on

arrangements to enable the exchange of traffic

across different networks.

Two types of arrangements evolved.

The first type is transit, a paid connection

through a network to the parts of the internet

not available via a direct connection. Like

internet access purchased by end users, a

transit provider can link (directly or indirectly)

to all other networks on the internet, and their

associated customers.

The second type is direct interconnection

whereby two parties connect to one another,

but only for the purposes of exchanging traffic

destined for their respective networks (as

opposed to all of the possible end points on

the internet that can be provided by transit).

Interconnection may be settlement-free or

paid. In settlement-free interconnection, based

on a mutual benefit, no money is exchanged

between the parties. This was the basis of all

interconnection in the early days of the internet,

and is still the most common arrangement.

In these cases, there could be vast distances

between content and the user requesting it.

Therefore, as the internet grew and became

more international, some of these arrangements

became less practical for a number of reasons.

First, building and maintaining international

networks is costly, as it involves laying infrastructure

over long distances - overland and underwater.

In some cases, this connectivity doesn’t exist

and would need to be built from scratch. In other

cases, delivering traffic to a location will require

passing through multiple networks.

Moving content closer

90%

$75B

of internet traffic destined for

consumers is carried by CDNs

is invested annually into infrastructure

by online service providers to bring

content closer to consumers

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Second, remote content must travel through

more routers (the internet’s switches) on its

journey to the consumer. Each one of these

routers has a chance of being congested by

high traffic loads, which can lead to packet

loss

- the discarding of certain elements of

the data being transmitted. Packet loss above a

certain level results in a poor experience for the

consumer, for example in the form of buffering

and/or degraded video quality. In extreme cases

it may render the service unusable.

Third, geographic distance leads to delay

(latency) in the time it takes a packet of data to

reach the consumer. This is less of an issue for

video streaming (where there is less interaction

and the data flows steadily

down to the

consumer). However, it can significantly degrade

other applications. Loading a web page may

involve multiple round trips for data between

the consumer and the server with the content.

The solution to these problems is to move

a copy of the content in question closer to the

consumer, using a distributed network of local

servers, i.e. a Content Delivery Network (CDN).

This reduces the geographic distance content

must be carried, shrinking an ISP’s costs.

CDNs also minimize the number of routers and

networks content must travel through, minimizing

opportunities for congestion.

According to UK regulator Ofcom:

The local

delivery of content can result in better delivery

times to the consumer, which may translate to

a better quality of experience, and so is often

a preferred option for content providers.

This

approach further reduces transit or backhaul

connectivity costs,

and can also improve the

customer experience by reducing the

likelihood of data congestion in these parts

of the network.

Many companies now operate such CDNs.

Large content providers operate their own

in-house CDNs. Smaller content providers make

use of independent CDNs such as Akamai, Fastly

and Lumen (formerly Level 3). Using a CDN

provides the ability to quickly and efficiently

deliver traffic across the internet and is an

important enabler of market entry and innovation.

CDNs are very widely and heavily used, and

one study found they carry approximately 90%

of internet traffic destined for consumers.

Analysys Mason estimate that in the period

2014-17 online service providers invested

$75bn annually in infrastructure that brings

content closer to consumers.

These

investments reduce costs for ISPs.

Netflix initially relied on independent

CDNs, but today operates its own CDN,

called ‘Open Connect’.

A data packet is a small unit of data.

Strictly speaking, video is served in chunks, which fill a buffer

at the user end. As the user watches the video and depletes

the buffer, a new chunk will be sent. However, this is all

invisible to the user.

These are costs for the long-haul distribution of traffic,

which may be borne by an ISP and/or a content provider

Ofcom, Connected Nations 2016, 16 December 2016

Craig Labovitz, Internet Traffic 2009-2019, 26 February

2019. A range of 60-80% was reported for Brazil in 2018:

NIC.br, A importância dos Sistemas Autônomos e dos

Internet Exchanges/PTTs, 23 August 2018. An earlier

Ofcom study reported 83% of fixed network traffic came

from CDNs in 2017. Ofcom, Connected Nations 2017

- Data analysis, 15 December 2017.

Analysys Mason, Infrastructure Investment by online service

providers, December 2018.

Long way round

Data would be forced

to travel through many

routers and networks

Lost in transit

Congestion would lead

to ‘packet loss’, causing

quality to degrade

Traffic jams

Each router could

be congested by

high traffic volumes

SINGLE SOURCE

When a user requested

video content, it was served

from a single location

CONGESTED DELIVERY

This server may have been a huge

geographical distance from the user

and traffic growth would mean

expanding this long distance capacity

HUGE EXPENSE

Securing global connectivity

would come at a high cost

to Internet Service Providers

(ISPs) who would have to

purchase or build these

backhaul connections

A Netflix briefing paper

2021

A cooperative approach

to content delivery

The long haul

Why the early structure of the internet could not scale

to meet the demands of today’s internet usage

A Netflix briefing paper

2021

A cooperative approach

to content delivery

1. 2

$ B

The amount of money

Open Connect saved

ISPs in 2020

A Netflix briefing paper

2021

A cooperative approach

to content delivery

As of early 2021, Netflix has over 200 million

paying customers around the world, using tens

of thousands of combinations of device and

network configurations. Open Connect is a key

part of ensuring that these customers can reliably

access high-quality video streams as efficiently

as possible. Netflix also uses other techniques,

such as sophisticated video compression and

file versioning to reduce the size of video files,

which we discuss later.

Open Connect CDN is a combination of local

servers (called Open Connect Appliances -

OCAs) and backbone infrastructure. Since the

launch of Open Connect in 2011, Netflix has

spent over US$1B to develop and deploy over

14,000 OCAs across 142 countries.

Description of

Netflix Open Connect

Illustration of Open Connect architecture

Users

Origin

server

Wider internet

OCA content refreshes

every night

OCAs can sit at IXPs or

within ISP networks

Broadband

access

ISP

OCA

IXP

OCA

A Netflix briefing paper

2021

A cooperative approach

to content delivery

In order to bring content as close as possible

to the consumer, OCAs are either deployed

within ISP networks or at public Internet

Exchange Points (IXPs), physical locations

where many networks come together to

exchange traffic. Each OCA deployment stores

nearly all of the Netflix catalogue, including each

title in the various formats and video qualities.

Content is refreshed overnight as new content

is released or becomes more popular.

Open Connect will directly interconnect with

any ISP to hand off traffic at one or multiple IXPs.

The interconnection predominantly supports

the flow of traffic from Netflix to the ISP, with

each side upgrading the capacity as needed

to support user requests for content. Unlike

interconnection among ISPs, where traffic is

exchanged in both directions, traffic is primarily

content destined for the ISP’s users that have

requested it, so it is flowing in one direction.

OCAs within an ISP’s network are provided

by Netflix to the ISP free of charge on a

non-discriminatory basis. In most cases,

the ISP takes full ownership and control of the

equipment. Installing an OCA within the ISP’s

network allows the ISP to place it wherever in its

network is most beneficial, avoiding costs that

might otherwise be incurred fetching content

over long distances to bring it into its network.

To date, over 1,000 ISPs have taken ownership

of and installed OCAs. For our ISP partners that

have opted to deploy servers within their own

networks, the cost savings for 2020 is estimated

to be approximately $1.2B.

If an OCA is hosted at an IXP, then Netflix

retains ownership and pays for its own power

consumption, colocation fees, cross-connect

fees and so on. Netflix has placed OCAs in more

than 80 IXP locations in over 25 countries,

where Netflix will peer with any ISP.

This investment by Netflix can have spillover

benefits when a content provider like Netflix

builds out its CDN to an IXP, ISPs and other

content providers may be encouraged to also

have a presence at that IXP. Multiple content

providers and ISPs coming together at a single

IXP location can generate significant efficiencies

for all concerned.

Cost savings calculation

In turn the price per unit of traffic

depends upon the cost of leasing

long haul capacity.

The cost savings estimate reflects

the savings that ISPs that take OCAs

within their networks realize as a result

of significant localization via the Open

Connect Appliances and the resulting

reduction in long haul internet backbone

costs. It is important to note that even

partners that are unable or unwilling

to take OCAs within their network still

realize savings as a result of localization

of content within IX sites. Without the

Content Delivery Network (CDN) content

would be handed off at the source of the

data (a handful of cloud storage regions)

and the ISP would then carry the content

to their end users.

Volume

of traffic

x x

Distance

traveled

Price per

unit of traffic

Estimate of costs per year

100

Netflix headquarters

in California, USA

KEY

Netflix OCAs

Located within public

Internet Exchange Points

OCAs gifted to Internet

Exchange Points

(ISPs)

Number of content

caches per server

Bringing content closer

Netflix’s investment in a new world of content delivery

OPEN CONNECT

Netflix now stores its content close to every member

- no matter where in the world they are.

Open Connect is Netflix’s Content Delivery Network.

It’s made of more than 14,000 Open Connect

Appliances (OCAs) spread across 142 countries.

A Netflix briefing paper

2021

Note: Data as of 2020

A cooperative approach

to content delivery

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Cost savings

Having content delivered by Open Connect

provides various cost savings to ISPs.

First, they avoid ‘transit charges’. While many

ISPs will freely exchange traffic with other ISPs

in their own region by directly interconnecting

with each other,

to exchange traffic with more

distant networks they generally rely on (and pay)

transit providers. Further, they have to ‘meet’ the

transit provider at an overseas location if transit is

unavailable in their own country. In this case, the

ISP will need to pay not only the transit charge,

but also the costs of expensive international

backhaul capacity to reach the transit provider.

However, Open Connect allows ISPs to avoid

these costs, by passing the traffic directly to the

ISP rather than via a transit link.

According to TeleGeography:

While the

increase in broadband users and access rates

will continue to drive traffic growth in access

networks, much of this growth may be managed

locally within a network and may not lead to

proportional increases in traffic on international

links. Thus, CDNs and caching will continue to

have a localizing effect on traffic patterns and

dampen international internet traffic growth.

French ISP Altice has noted that embedded

Open Connect OCAs have:

Advantages both

for the ISP and Netflix, minimizing the Internet

traffic, while improving the quality experienced

by customers, respectively.

Benefits for ISPs

from Open Connect

Second, if an ISP hosts multiple OCAs, it

also avoids some cost for their own core network

capacity. Traffic can be delivered to the consumer

from a nearby server, rather than having to travel

across the ISP’s core. UK broadband provider

Sky has cited CDN caching as a key source of

cost saving as video demand grows.

Case study: Brisanet

Brisanet is a regional ISP that operates

in the Northeast region of Brazil. Netflix

traffic represents about one fifth of

Brisanet total traffic.

Localizing the traffic has helped

the company to significantly reduce

transport traffic within Brisanet’s

network and thus reduces costs for

augmenting transport capacity.

Localizing traffic has also improved

the quality of experience for consumers.

After installing Open Connect servers,

Brisanet has received very positive

feedback from their customers via

their call center.

See the appendix for a more detailed discussion.

TeleGeography, The State of the Network, February 2020.

Altice Labs, InnovAction #2, January 2018.

Sky, Sky Core Network - Positioning for Massive Video

Demand, 20 Feb 2018.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Netflix Open Connect is uniquely able to help

ISPs take advantage of these benefits because

we have a defined, limited catalogue of content.

Netflix is able to better optimize the upload

content to OCAs during off-peak hours, which

avoids burdening the ISP or other parts of the

internet during periods of the day where there is

the highest usage. This means the connectivity

needed to fill these servers is often without any

cost to the ISP, as network capacity is built (or

acquired) for peak capacity. Netflix is uniquely

able to pre-position copies of all content in

advance. Many content providers store a copy

of a file only after it has been requested by

a user, which may occur during peak hours.

Furthermore, because the volume of Netflix

traffic is limited, an OCA deployment can offload

up to 100% of traffic from a single location.

Increased revenue from consumers

Aside from these immediate cost benefits,

the improved quality of experience for the

consumer has benefits for ISPs. Consumers who

are happy with their experience of the services

they receive over their broadband connection

are unlikely to change their broadband supplier.

By preventing buffering and instability, CDNs

help broadband networks retain their customers.

A survey of broadband users found that slow

speed/buffering and poor stability were the two

main reasons that would influence broadband

consumers to change their supplier.

Network operators recognise the quality

and cost benefits of CDNs. According to UK

ISP TalkTalk:

Caching helps manage capacity,

customer experience and costs … Netflix, for

example, is located at over 60 locations around

the UK, avoiding bandwidth costs in the core

network and serving content to end customers

with lower latency response times.

GSMA, the global association of mobile

operators, has said:

By having content closer

to the subscribers (using CDNs), the expected

delays are shorter, and this means better quality

of experience. Also, by having visibility on the

services and the locations where those services

are being used, the operator can spend less

on network optimisation, prioritising the most

important areas of high-value services. Users

with better QoE [Quality of Experience] directly

translates to less churn.

In addition to these advantages, CDNs are

more resilient, since if one element fails, traffic

can be redirected to other servers, reducing the

risk of downtime.

Huawei / Strategy Analytics, Unlocking Value for Residential

Broadband Services with Quality Broadband Network,

October 2017.

TalkTalk, Capital Markets Day, 5 March 2019.

GSMA, América Móvil and VIAVI Solutions – Creating

Business Value through Content Delivery and Analytics, 17

December 2019. Text in brackets added for purposes of

clarification.

Case study: Telecentro

Telecentro is an ISP in Argentina.

Thanks to Open Connect, the company

was able to significantly reduce its costs

for international transit and improve

performance. Because of that, their

customer base grew faster as they

were perceived as a very high quality

broadband ISP in the market.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Greater control

Open Connect provides ISPs with unparalleled

control over how they receive Netflix traffic.

Video-on-demand traffic has a largely predictable

daily pattern, rising to a maximum in the evening.

This allows content providers and ISPs to better

predict and account for peak traffic. ISPs have full

control over where they receive traffic, whether

at one of the many IXP locations where Netflix is

present or through placement of OCAs within

their network. ISPs can fully determine which

of their customers are served from which OCA,

enabling better network planning and resilience.

ISPs are also able to determine when the

content on OCAs is refreshed with the latest

Netflix films and series. This allows them to

ensure content is refreshed at off-peak times,

so it does not contribute to peak capacity needs.

This can help to minimize costs and provide

greater network stability.

Encoding

Open Connect is not the only tool Netflix uses

to ensure a good user experience. It also seeks

to use the least amount of data to deliver a

given picture quality.

Delivering quality video to Netflix’s members

without using unnecessary bandwidth allows

a good experience even on intermittent

connections or those with limited bandwidth.

But efficient encoding is applied to all Netflix’s

streams, and this has a benefit for ISPs too,

since it minimises the traffic flowing across

their networks.

There are also benefits to members beyond

just picture quality. If less of a household’s

broadband capacity or wifi capacity is used for

a given Netflix stream, then the freed capacity

can instead be used for other applications.

There are two aspects to creating more

efficient video streams. The first is that Netflix

encodes multiple versions of the same video

file, customised to the capabilities of different

devices (including legacy ones) and the available

bandwidth on a member’s internet connection.

The multiple encoding versions adapt to the

device and available bandwidth automatically,

allowing the resolution to go up depending on

what a device is capable of and/or congestion

on the broadband connections to viewers’

homes. This is efficient since there is no benefit

to sending a heavy, ultra high definition video

stream to a device with a lower resolution screen

or an internet connection that can not handle it.

The second is a drive to improve video

compression. Successive generations of video

compression formats allow the same quality of

visual image to be delivered with less and less

bandwidth. Netflix invests continually to deploy

the most effective compression. It also optimises

compression for different content types - using

less data to deliver less complex video. For

example, a video of a news anchor can be encoded

using significantly less bandwidth than a car chase.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

As a result of these efforts, Netflix streams are

increasingly efficient. The number of hours a Netflix

member can stream per GB of data has increased

by more than 200%

over the last five years.

Netflix is also working alongside a number

of companies including Samsung, Intel, Cisco,

and Hulu in the Alliance for Open Media, a

collaborative effort to offer open, royalty-free

and interoperable solutions to make media

technology more efficient. These solutions seek

to provide superior quality for all users, on all

devices, and on all platforms. The Alliance’s

efforts have resulted in the launch of video

encoding technologies that have dramatically

improved the efficiency of high quality video

streaming across the industry.

Conclusion

These encoding improvements and deployment

of Open Connect offer substantial benefits to

ISPs, including cost savings, the stability of

services they provide, and increased revenues.

ISPs may see significant reductions in transit

costs from moving content closer and ISPs who

deploy servers in multiple locations may also

see a reduction in the need for infrastructure

to connect those locations. Maintaining

international connectivity is costly as it involves

infrastructure that spans long distances. Long

distance connections have more opportunity to

become congested by high traffic loads, which

can lead to a poor experience for the consumer.

Delivery through Open Connect “opens up

the pipes” of an ISP to reduce the chances of

congestion impacting Netflix or other application

quality. The increased quality of video delivered

via a CDN can lead to reduced churn for the ISP.

In the past five years, Netflix’s encoding has more

than doubled the data efficiency of streaming

Based on the latest encoding profile.31

50 10 15 20

Hours of streaming per 4GB of data

2019 AL1-VP9

2015 CBE1-AVCMain

2018 AL1-AVCHigh

2020 CE4-DO-HEVC

2020 AV1

It once took 4GB to

stream 11 hours of content

Netflix

members now

stream 25 hours

for 4GB

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Threats to

the ecosystem

Understanding

traffic charges

The case for traffic charges

Dangers of traffic charges

A Netflix briefing paper

2021

A cooperative approach

to content delivery

In recent years, one of the key changes to which the

internet has adapted has been the increase in video traffic.

Improving access networks, rapidly falling

costs of transmission and substantial investment

in CDNs by video providers now enable

subscribers to reliably watch high quality

video on a range of devices.

As we have seen, the growth in video

(supported by CDNs) has had benefits for

consumers, content providers, ISPs and society

as a whole.

However, these benefits rely on a somewhat

unstable equilibrium of mutual cooperation to

invest in efficient delivery. There is no requirement

to move content closer to consumers. An

uncongested long haul connection can still

deliver high quality video.

Thus a relatively

small disincentive to bringing content closer to

Latency may be increased if there is an international leg, but

this is relatively unimportant for video. However, if the transit

link is congested, then there is a risk of packet loss and lower

quality video.

ETNO, ETNO paper on Contribution to WCIT,

7 September 2012.

consumers may prompt a content provider to stay

further away. However, this would cause additional

costs for the ISP, and increase the likelihood of

congestion on back-haul networks.

It is important not to inadvertently create

incentives that would undo the cooperative

approach that has led to the investments in

moving content closer by way of Content Delivery

Networks. Despite these risks, some governments

and ISPs have considered charging content

providers for access to networks. This is likely

to harm consumers, through double charges for

connectivity, less competition, and ultimately

poorer performance. In this chapter we consider

the (flawed) case made for traffic charges,

and then turn to the damage they may cause.

The case for traffic charges

The most frequent argument in favor of traffic

charges is that ISPs are in urgent need of further

funds for investment to support increasing traffic

volumes and to fund wider broadband deployment.

Past claims that traffic charges are

essential have proven unfounded

These arguments are not new. For example,

in 2012, ETNO, the European association of

incumbent telecoms operators, argued that:

Today there is a huge disproportion amongst

revenues and a clear shift of value towards

players (Over the Top players ‐ OTT) who are

not contributing to network investment. Traffic

and revenue flows need to be realigned in

order to assure the economic viability of

infrastructure investment and the sustainability

of the whole ecosystem.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

ETNO went on to propose the end of

settlement-free interconnection (the collaborative

exchange of internet traffic between ISPs and

content providers) and the introduction of a

‘sending party network pays’ principle - that is,

traffic charges. This proposal was not adopted,

but even so the internet continued to function

perfectly well - traffic charges were not necessary

for investment or sustainability.

In the US there has been a ‘natural experiment’

in the significance of charges to content providers

for investment. When Charter Communications

(a US cable operator) merged with Time Warner

Cable and Bright House Networks in 2016, the

merger was subject to a number of conditions,

including an obligation to offer settlement free

interconnection with video providers.

While this restriction has since been removed,

we might have expected to see a degradation

in the performance of Charter’s broadband,

relative to other ISPs that were not restricted to

settlement-free interconnection over the time

period in which payments were prohibited. In

fact, Charter has had one of the best performing

cable broadband networks in the US, according

to the FCC

and the ISP Speed Index.

The lack

of paid interconnection (charges by an ISP to a

content provider for a direct connection) has not

fed through to poor performance for consumers.

Thus both in the case of ETNO members

and of Charter, traffic charges have not been

necessary to fund improved broadband.

Indeed, in recent years there has been a surge

in investment in fibre broadband globally. Much

of this has been purely commercial, and has not

needed traffic charges as a justification.

Traffic charges are a poor

mechanism to support investment

Even if new sources of funds were necessary to

support broadband investment, traffic charges

would be a poor mechanism. This is because any

payments received by ISPs would be ‘untied’

- that is, there would be no requirement for the

funds to be spent on broadband infrastructure.

ISPs may have many uses for their cash - the

purchase of sports rights for an IPTV service,

acquisitions, returning cash to shareholders and

so on. These are all perfectly legitimate - but

funds received from traffic charges may be spent

on these, rather than on the broadband upgrade

that was the purported purpose.

Better interventions are

available to support widespread

broadband deployment

While traffic charges are not an appropriate

tool to support network investment, we are not

suggesting there is never a case for external

support. For example, there is a clear social

case for supporting the provision of good-

quality broadband in rural areas where it may be

commercially unviable to deploy. However, there

are many available interventions to support rural

broadband, and indeed many countries already

have such interventions in place. For example,

in the United States some states are providing

funding to support broadband deployment in

unserved and underserved areas through grant

programs that fund a portion of the cost of

deployment in these communities.

Traffic charges would be an untargeted and

inefficient way to support rural broadband.

They also carry significant dangers.

US Department of Justice, Justice Department Allows

Charter’s Acquisition of Time Warner Cable and Bright

House Networks to Proceed with Conditions, 25 April 2016.

FCC, Ninth Measuring Broadband America Fixed

Broadband Report, 3 August 2020

https://ispspeedindex.netflix.net/country/us

The Pew Charitable Trusts, How States Are Expanding

Broadband Access, February 2020

200

The number of hours a Netflix

member can stream per GB

of data has increased over

the last five years

A Netflix briefing paper

2021

A cooperative approach

to content delivery

A Netflix briefing paper

2021

A cooperative approach

to content delivery

The termination monopoly

The termination of traffic - final delivery to a

consumer - is the last great telecom monopoly.

If a given user requests a video stream while

connected to a particular ISP, the video provider

has no option but to send that stream via that ISP.

This remains true even if the user has access

to other networks. It is clearly not possible for

the video provider to request that the user

disconnect from their home wifi and switch to

4G, to watch the video in question.

The danger of termination monopolies is one

of the reasons regulators have been so active in

the regulation of voice termination rates on both

fixed and mobile networks over many decades.

Without this regulation, there would have been

a real risk of excessive pricing by telcos.

The termination monopoly for internet traffic

is relatively benign if a consumer’s ability to

reach any content of their choice is not impacted

by that ISP. However, if ISPs charge for the

termination of internet traffic then consumers’

ability to reach content may be restricted to

those content providers that are able to pay.

Furthermore, restricting consumers to only

a subset of content, not a broad spectrum of

content they might not otherwise encounter,

deprives society of a rich content portfolio.

The risk of a termination monopoly is

exacerbated in markets where consumers

face challenges switching ISPs - either due to

limited competition, long-term contracts, or high

switching costs. But even in markets where there

is competition among ISPs, a content provider can

only reach consumers through their chosen ISP.

Dangers of traffic charges

Theoretical alternate routes to the consumer

ISP

Traffic from content

provider via transit

Traffic from content

provider via peering

User

Content

provider

Transit

provider

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Alternate routes to consumer

often not realistic in practice

It is sometimes argued

that such termination

monopolies are unproblematic because content

providers have various routes to a given ISP.

For example, the content provider might have

a direct connection to the ISP via peering

and

an indirect connection via transit, as shown on

the previous page. (Alternatively, they might

have CDN servers on-net - that is within the

ISP - and transit).

However, this is often more the illusion

of choice than the reality. The ISP ultimately

controls all of these routes into its network

and can modulate the quality and potentially

the pricing of both these routes. (The ISP

might charge the transit provider for paid

interconnection, for example). Thus, this

apparent choice of route for the content

provider invariably leads back to the access

provider, who is at best competing with itself.

Further, transit links are unlikely to have the

capacity to absorb traffic redirected from direct

interconnection or a CDN. Ofcom (in 2017) found

that for UK fixed networks, transit represented

just 5% of their traffic,

and transit capacity is

presumably provisioned accordingly.

As we have seen, Netflix alone might represent

11% of an ISP’s traffic. Thus if Netflix were to

switch its traffic to transit from Open Connect,

16% of the relevant ISP’s traffic would be flowing

through circuits designed for 5%. (Even this

assumes that routes are available through all

the ISPs’ transit connections, which is far from

certain). This would lead to massive congestion

and consequently a catastrophic degradation

of quality for Netflix’s services. Netflix’s

competitors, likely still delivered via CDNs within

the ISP’s network, would be largely unaffected.

This congestion would also create some

problems for the ISP, since other services arriving

via transit would also be affected.

See, for instance, Laure Jaunaux & Marc Lebourges

(Orange Regulatory Department). Externalities between

on-line contents drive telecom operators’ incentives to

provide quality open internet through neutral network,

14 January 2019.

Peering refers to a direct connection to exchange traffic

between two networks, generally at zero cost.

Ofcom, Connected Nations 2017

- Data analysis, 15 December 2017.

1000 Mbps

200 films

100 Mbps

20 films

25 Mbps

5 films

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Fast times

As broadband speeds get faster, Netflix streams take up less and less

of a member’s purchased bandwidth. Netflix recommendsjust 5 Mbps

to stream a high-definition movie on Netflix on a fixed line connection.

That means a Netflix member with a high speed broadband connection

has enough bandwidth to stream multiple HD videos at the same time

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Potential for leveraging

into video markets

The threat of this termination monopoly is not

simply a matter of the price paid for termination.

Many ISPs compete directly with streaming video

services, either because they are themselves

a cable TV operator, or a telco with an IPTV

offering. By 2024, 45% of Western European

households are expected to take TV from a

telco or cable operator.

ETNO has noted

that “Operators supplying traditional pay-TV

within Western Europe continue facing strong

competitive pressures from OTT services

[such as Netflix and Amazon].”

Thus for any ISP with its own TV service, any

degradation of quality for other video providers

strengthens the competitive position of that TV

service. This risks providing an incentive to take

advantage of the terminating monopoly in a way

that is detrimental to consumers.

Potential for double charging

ISPs seeking to charge content providers for

the carriage of traffic have almost universally

already charged their consumers for carriage

of that same traffic. Both mobile and fixed

broadband products include data allowances

(often ‘unlimited’ for fixed). Consumers have paid

for these data allowances on the understanding

that it will allow them to reach the entire internet,

not just those sites that have agreed to pay to

be distributed by their ISP. Further, if consumers

have paid for traffic, to also charge content

providers would appear to be a double-charge.

This problem is best illustrated when the

speeds required for the content being delivered

is well under the speeds paid for by consumers.

The average global fixed line broadband speed

is 87.84 Mbps

and this is much more than is

needed to watch video. A Netflix member should

theoretically be able to watch 17 simultaneous

HD streams (of 5 Mbps each) with a connection

of this speed.

Perverse incentives

The only way for an ISP to force a content

provider to pay for traffic charges is to ensure

congestion otherwise limits an ISP’s customers

from receiving requested traffic from the content

provider. This congestion represents powerful

leverage for the ISP to force the content provider

to accept the demanded cost to connect directly,

and thus the ISP has little incentive to remedy

the congestion.

This potential leverage has caused concern

to regulators in the context of ISP mergers,

and, in the past, ISPs have been known to use

congestion in this way.

Digital TV Europe, Rise of IPTV to benefit Telcos,

23 May 2019.

ETNO, The state of digital communications 2020,

28 January 2020.

For a detailed discussion of these issues, see David Evans,

Comcast’s Acquisition of Time Warner Cable Would Result

in an Economically Significant Increase in the Magnitude of

Terminating Access Fees for Online Video Distributors,

6 April 2015.

Speed Test Global Index, https://www.speedtest.net/global-

index, Accessed on 3 December 2020. The global average

download speed for October 2020 was 87.84 Mbps.

See for example ¶550 onwards of European Commission,

Case M.7000 - LIBERTY GLOBAL / ZIGGO, 30 May 2018

A number of such interconnection disputes are discussed

in Daniel A. Lyons, “An Antitrust-Informed Approach to

Regulating Internet Interconnection.” Journal of Science

& Technology Law 24, no.2 (2018): 229-276

A Netflix briefing paper

2021

A cooperative approach

to content delivery

If an ISP does impose charges for

interconnection or on-net CDNs, then this

gives content providers two choices: pay those

charges, or deliver their traffic via offshore

methods (e.g. via transit or non-localised

interconnection). Transit prices have had a

long-run trend of brisk decline, so in financial

terms this can be an attractive option for content

providers. This would result in backbone and

transit links being used for content that can be

more efficiently delivered, both in terms of cost

and capacity, by means of local content servers.

While traffic charges by an ISP may force

such an outcome, it is clearly suboptimal. For the

ISP itself, it will likely mean that it is now paying

to receive the traffic, since it may need to pay

transit charges to the relevant transit provider.

Moreover, if the ISP is buying its transit at a

remote location, then it will need to invest in

international backhaul capacity also.

Avoiding such costs is one of the key reasons

more than 7,000 ISPs around the world use

Open Connect to receive traffic. Larger ISPs

may tolerate these increased costs in order to

force through traffic charges, but there is still

risk to consumers.

Consumers invariably lose in this situation.

They will be exposed to the risk of congestion in

transit links, with possible degradation of their

user experience. Increased latency may also

cause problems (though less so for streamed

video, which is not latency-sensitive). There is

also a risk that consumers will lose access to

the plurality of voices and applications that have

enabled the internet to thrive.

Transit price CAGR, 2017-20

TeleGeography, Global IP Transit Prices Keep

Doing What They Do Best, 8 September 2020.

-40%

-50%

-30 -20 -10 0

Miami

Johannesburg

Sao Paulo

London

Singapore

Sydney

Mumbai

CAGR

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Conclusion

A cooperative model is win-win for consumers.

At the heart of the internet’s value is the fact that it

is an adaptable network of networks which enables

the communication and sharing of information and

content between people, irrespective of location.

In recent years, one of the key changes to which

the internet has adapted has been the increase in

video content. In order to support this increase in

traffic, content providers like Netflix have worked

hand in hand with ISPs.

In Netflix’s case we invest significantly in

content, which drives demand for broadband

internet access. We invest in delivery infrastructure

(OCAs and backhaul infrastructure) and

compression technology. In both cases, we have

not only innovated on existing industry standards,

but also actively participated in standardization

activities to ensure constantly improving efficiency

across the industry as a whole.

We have made these investments in

delivery because we want to give our members

the best experience possible no matter what

the available bandwidth is or the capability

of their viewing device.

And our investments have also reaped

significant benefits for ISPs - increased

revenues, greater broadband takeup - and for

many ISPs Open Connect has enabled them to

avoid the costs of expensive international transit.

It has also generated benefits to society. The

availability of high quality video gives consumers

a reason to go online in the first place and to

upgrade their broadband connections, promoting

greater connectivity.

Looking to the future we are keen to continue

to work in partnership with ISPs in this way to

ensure that content is delivered to our members

in the most efficient way possible. However

a cooperative model will only be sustained

if content providers and ISPs continue to

recognize the complementary nature of networks

and content. In addition, if governments

acknowledge the important role that content

providers play in stimulating demand for

broadband services and establish supporting

policy frameworks which restrict the imposition

of charges by ISPs on content providers.

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Glossary of terms

CAGR

Compound Annual Growth Rate

Cache

A temporary local copy of information that

originated elsewhere. Thus for CDNs, a copy

of files to be delivered to consumers, stored

in a local server

CDN (Content Delivery Network)

A distributed system of servers, designed to

enable the efficient and reliable distribution of

content over the internet

CP (content provider)

An internet business whose focus is delivering

content (rather than - say - e-commerce)

to consumers. Netflix, CNN and YouTube

are examples

FBB

Fixed broadband

Hop

One step in a packet of data’s journey

across multiple servers

ISP (Internet Service Provider)

A company providing internet connectivity

to consumers (consumers or businesses). May

provide fixed broadband, mobile data or both

IXP (Internet Exchange Point)

A location where many networks meet

to exchange traffic, avoiding the need

for multiple bilateral connections

Latency

The lag between a packet of data being

sent and reaching its destination

OTT (Over The Top)

Describes services delivered over another

network without being integrated with it.

YouTube, Facebook and Netflix are examples

(since they are not provided by telcos operating

broadband networks)

Packet loss

When a router is sent more data than it can

handle, it discards a certain amount of data.

This is known as packet loss. Typically the

data in question will then be requested again

from the source server

Router

A switch on the internet, that receives packets

of data and sends them onwards down the

appropriate link

Server

A computer that stores and

transmits content

TCP/IP

Transmission control protocol and internet

protocol. The two foundational standards for

data transmission that underpin the internet

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Appendix

The Internet (TCP/IP) protocol enables easy physical

interconnection between two networks that both

use it. However, there is also a commercial aspect

to interconnect - does money flow between the two

parties and what extent of connectivity is provided?

There are two primary categories of

interconnection, namely transit and direct

interconnection (with some variety under

each heading). We discuss these below.

Transit

and peering

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Transit

Transit is typically used by an ISP (or content

provider - CP) to reach the parts of the internet

not available to it via other types of connection,

such as interconnection. The ISP buys transit

from a transit provider, who in turn links (directly

or indirectly) to all other networks on the internet,

and their associated consumers.

For destination ISPs and CPs on its own

network, the transit provider passes the traffic

on directly. For other destinations, it will connect

with other transit providers, who will in turn pass

the traffic on to the destination in question. Since

essentially all ISPs and CPs are connected to at

least one transit provider, connection via such

providers enables universal connectivity.

While transit provides universality, it has two

major disadvantages. Firstly, it comes at a cost.

The ISP or CP must pay in relation to the amount

of traffic carried via the transit link.

Secondly, it may lead to inefficient traffic

routing. An ISP may have to procure transit from

a provider that is not based in their own country,

but (say) in the US. If so, all traffic sent by the ISP

will have to travel to the US, even if it is destined

for a consumer in the ISPs’ own country. This is

known as traffic tromboning. It is expensive, since

both the sending and the receiving provider must

pay for pricey international capacity. It is also

technically poor, since the long journey introduces

latency and the potential for packet loss.

It is also possible to buy partial transit, where the transit

provider offers connectivity only to certain destinations

Transit is typically priced on a 95th percentile basis - that is,

based on ‘near peak’ bandwidth demand, but setting aside

the busiest 5% of periods

While tromboning is becoming rare in larger internet markets,

it remains a significant problem in some smaller markets.

See, for instance, UN ESCAP, An In-Depth Study of the

Asia-Pacific Information Superhighway in CLMV Countries,

March 2020

Illustration of transit

Illustration of peering

Peering

ISP

Broadband

access

ISP

Broadband

access

ISP

Broadband

access

ISP

Broadband

access

Content

provider

Transit

provider

Cash flow

Transit

provider

Content

provider

User

User User

A Netflix briefing paper

2021

A cooperative approach

to content delivery

A Netflix briefing paper

2021

A cooperative approach

to content delivery

Direct

interconnection

The main alternative to transit is direct

interconnection through peering. When two

networks peer, they connect, but only for traffic

destined for their respective networks. For

example, ISPs X and Y might peer, and the

traffic flowing over the link would be any traffic

travelling from ISP X’s consumers to ISP Y’s

consumers, and vice versa.

ISPs may peer, but so too can ISPs and

content providers. In this scenario the traffic

flowing would primarily be the content destined

for the ISP’s users.

There are a range of types of interconnection.

It may be public or private. Public interconnection

takes place via an Internet Exchange Point (IXP),

a physical location where many networks come

together to exchange traffic. This is efficient for

connecting with many different peers. Private

interconnection takes place via a direct physical

connection between two networks. This may be

lower cost and more reliable if a large volume of

traffic is being exchanged.

Peering may also be settlement-free or paid.

In settlement-free interconnection, no money

is exchanged between the peers. This was the

basis of all interconnect in the early days of the

internet, and remains very common. Generally

interconnection links are mutually beneficial,

and so both parties gain even if they are unpaid.

Further, if there is no payment then contracts can

be far simpler. Indeed, interconnection based on

a handshake (i.e. without a written contract) is

very common.

That said, networks often have certain criteria

that they apply to certain peers, to avoid being

gamed. For example, a network with expensive

international connectivity might be reluctant to

peer with a purely local network, since it would

pick up the lion’s share of the cost of transmitting

traffic between the two networks’ customers.

Sometimes networks only offer

interconnection if traffic flows are roughly

balanced - that is, there is a similar volume of

traffic flowing from peer X to peer Y as is flowing

in the opposite direction. This does ensure that

the two parties are imposing similar traffic costs

on each other. However, this is not necessary for

settlement free interconnection to be merited.

Asymmetric traffic may create value for both

parties, even if costs are not balanced. For

instance, YouTube may send a lot of traffic to

a small ISP, but that ISP gets great value from

making YouTube available to its consumers

While settlement-free interconnection is

common, there is also paid interconnection,

where one party pays the other. This may arise

where one party gets much less value out of the

interconnection (and so requires compensation),

or where one party has leverage, and can simply

insist on payment.

‘Partial interconnection’ is also possible, where

peers only agree to connect certain subsets of

their customers, similar to partial transit