Internet Perspectives May 2002 Geoff Huston Chief Scientist, Internet Telstra.

Internet Perspectives

May 2002

Geoff HustonChief Scientist, Internet

Telstra

One View of the Internet “Moving from disruption to disruption”

Characterization of the Internet as a disruptive technology

Internet deployment is driven by a succession of disruptive events

Market survival is characterized by adaptation to rapidly shifting models in the wake of each disruptive technology

Market leadership is characterized by advance identification of disruptive events

Internet Disruption Events The so-called “Killer -Apps” of the Internet:

Email 1988

FTP 1990

The Web 1993

Portals 1996

Instant Messaging 1998

Napster 1999

Contenders for future K-As Mobility & Wireless E-identity services Music Distribution Video Distribution GRIDs Telephony services Appliances ??

BUT… The problem with this perspective on the

Internet is that leading market players are forced into making investment decisions on likely directions in disruptive environments. This predictive investment has a high risk There is little tolerance left for high risk options

in this sector Players are looking for a more conservative

approach to investment in this sector

Another View of the Internet Packet carriage medium

Packet architecture Address architecture Flow control protocol

The so-called ‘hourglass’ model of IP as a generalized adaptation layer between the communications medium and the application

‘Hourglass’ IP Objectives IP is not an end in itself

It enables applications to work across a diverse set of environments

IP is not a panacea for all communications requirements

Some applications demand a higher level of service than IP can effectively deliver

But IP has a market role by levering off three major assets:

potentially cheaper than alternatives flexible in that it makes few assumptions about

application behaviour scales into environments of high volume and high

speed

Technology Phases

1. Innovation2. Deployment and Adoption3. Exploitation

The Internet is now at the stage where the initial phases of innovation and early adopter models are completed.

The current phase is one of exploitation of the technology to lever advantage in other activity sectors

This phase is typically a commodity phase

Internet as a Commodity

Deregulated market with competing suppliers

Open market price information Uniform product with limited scope

for bundled value add as a product differentiator

Price is everything!

Commodity Markets

Prevailing market price determined by the second most efficient producer

Market share determined by Producer’s efficiency scale of production reliability

The Commodity Perspective of the Internet Each supplier must drive down its cost

of production in order to maintain market share

The drivers behind production cost for base level Internet services are Technology innovation Volume of production Stability of customer base Reliability of production Maturity of offering Complexity of offering

Looking Forward

Commodity markets are dominated by production efficiency

Producers are forced to create products that are: Cheaper to produce and operate Support a broader application base Support a larger, more diverse client

base

The Message to Providers Bundling and complex solutions (value-add)

should not be undertaken at the expense of base level efficiency of the product

Use simple architectures with basic functions Complexity costs rise disproportionately to size

Use extensible solutions Avoid excessive layering

Grand unified convergence is a myth Everything over Something can become a tragic

technology mistake Manage cost

The Message to Vendors

Fewer features in base components

Simple, stable platforms Component modularity Longer active lifecycles for

equipment Reliable and predictable operation

Specific Concerns

What technologies are of interest to carrier-based public Internet Service providers at present?

IP-based Technology Issues

Whats on our technology radar?

VPNs VOIP and ENUM QoS Identity technologies V6 DNS Zeroconf PnP

(in no particular order!)

Mobility Management

Architectures Multi-Provider last

mile access AAA and EAP TE and MPLS

IP-based Technology Issues Provider-Provisioned VPNs

Cross-product impacts with circuit-switched VPN technologies

Edge-to-edge overlay and shared secret alternatives Complexity issues of routing and topology maintenance,

QoS control, VPN stacking, network management and use metering

There is a suspicion that the value of the provider role in supporting PP-VPNS is greater than the incremental cost of supporting various levels of differentiation within the host network. This has yet to be conclusively proved.

There is also the belief that private data networks will continue to be valued as a premium offering by enterprise customers as a surrogate to effective distributed security solutions. This has yet to be contradicted.

IP-based Technology Issues VOIP and ENUM

Cross product impact with telephone revenue streams for value-added services

Ability to integrate enterprise private voice environments with the PSTN

Ability to create further value-added services that leverage telephone services

The concentration of interest is not so much in the carriage of voice over IP as the integration of switching control systems with IP-based distributed applications

IP-based Technology Issues QoS

Today - largely an enterprise approach to resource rationing

No large scale adoption within the public network environment

Issues with service management, metering, application interaction, inter-provider interaction, routing, complexity control, and viability of outcomes

It would be wonderful to charge disproportionately more for some packets. It would be a mistake if the costs associated with this functions are greater than the incremental revenue opportunities

IP-based Technology Issues Identity technologies

Most forms of e-commerce architectures rely on robust authentication and adequate privacy

Most forms of network abuse leverage off the weak level of authenticated identity that exists within the public IP environment

There is a view that a widely deployed trustable authentication service would enable wider adoption of online transactions across a larger client base

Shift the emphasis away from a trusted device to that of a trusted user of the device

The base technology is largely available – the regulatory and business models to support such a framework are still formative

IP-based Technology Issues V6

Any day now Really Trust me! The incremental deployment model of NAT technologies is

well-suited to the current collection of deployed applications and weak authentication

Large scale public deployments are increasingly based on private address space and NAT / ALG edges with limited capability provided to the end client

Requirements for stronger authentication and peer-to-peer applications drive a need for end-to-end coherency

But provider push is not enough – the actual driver is based in client pull, and to date the application base that drives client need for end-to-end coherency (V6) remains elusive

IP-based Technology Issues DNS

One of the most alarming cesspits of the public Internet infrastructure!

The distributed nature of the application requires strong authentication and security to operate with any degree of integrity

The distributed nature of the application ensures that this remains an elusive objective

A visible need to use advanced DNS technologies (DNSSEC, DNS IND, PKIX) to address the more overt weaknesses in this application

IP-based Technology Issues Zeroconf PnP

Broader deployment models encourage the use of self-configuring arrangements where a device establishes its address, routing and identity context using a model of trusted configuration agents

DHCP-based solutions have been effective in particular environments (dial and enterprise). Some further refinement of solutions appear necessary in the area of LAN-based connection services found in DSL, 802.11 and similar

IP-based Technology Issues Mobility

Recognition of high value solutions in the area of nomadic and roaming IP environments

Issues of differing technical solutions, differing transport characteristics, identity and location ambiguity, differing tariffs in the mobility domain, inter-provider roaming arrangements

Leverage of current mobile telephone infrastructure vs deployment of data-specific nomadic technologies

3G vs 802.11b,a,g

IP-based Technology Issues

Management Architectures Current element-by-element view of

management is ill-suited to an overall view of network integrity

If service management is an increasing topic of interest in enterprise overlays then management tools need to monitor the end-to-end delivered outcomes

IP-based Technology Issues Multi-Provider last mile access

Requirement to provide switching solutions that operate on policy-based constraints rather than header-based directives

A morass of competing technologies including various forms of PPP, L2TP VCs, LSPs coupled with policy-controlled aggregators

Highly complex switching environments with poor scaling properties

IP-based Technology Issues

AAA and EAP Desire to separate the access

mechanism from the billable end user Support of a variety of inter-provider

roaming arrangements that allow efficient use of access infrastructure

IP-based Technology Issues TE and MPLS

Intended to allow for more efficient use of network resources through managed load dispersal

Current routing-based approaches to TE suffer from uncontrolled feedback loops leading to network instability

The area of interest at present is not MPLS per se, but the ingress control systems which assign traffic into LSPs