Broadband Core Requirements

Imagine planning a party. You decide how many people to invite, choose a

venue with enough room for your guests and



plan for their arrival. But what do you do if you invite a hundred people and
instead three thousand turn up? This is exactly the dilemma facing today’s

service providers.

The Internet’s original core infrastructure was developed using scaled-up

legacy enterprise products. These legacy products are in stark contrast to

purpose-built Internet routers and do not offer the reliability, performance and

flexibility needed in the new broadband environment.

With the advent of broadband Internet and the new network services,

increasingly fast access technologies, including cable, fixed wireless and DSL,

are being deployed in the local loop. By 2005, Forrester Research predicts that

27 million European homes or 18 percent of the population will have broadband

access. Therefore, a robust and scaleable core infrastructure capable of

handling massive traffic loads is essential. Without a strengthened backbone in

place, many of the promises of broadband – such as high-speed access and

converged communication – are mere fantasies.

The Importance of Core Performance

In today’s competitive Internet market, a highly intelligent and speedy

infrastructure is crucial in order to meet ever-higher customer demands, but the

combined pressure of extra services and additional users poses threats to

service quality. For example, in order to effectively deliver advanced services,

routing technology sitting at the core of the Internet must be able to handle

traditional data traffic alongside new multi-service traffic, and maintain

forwarding performance during route fluctuations and congestion.

Legacy routers simply cannot keep pace with the demands of today’s

broadband networking environment. Why? First, their original design was centered

on the modest expectations of slower, less widespread Internet access and was

merely scaled up rather than redesigned. Second, they were designed for typical

loads rather than for exceptional conditions. The current high level of Internet

Protocol (IP) traffic over the in2frastructure was planned for within newer,

purpose-built Internet routers such as those from Juniper Networks.

Network surges are handled with speed and intelligence, using a combination

of electronic and optical network connections.

Core networks, with legacy routers, struggle to keep up with demand as

broadband permeates business and consumer premises. Designing networks around

normal operating conditions are not good enough to ensure the rock solid

stability the Internet requires. Packet forwarding performance goals are often

calculated by taking a mean of IP traffic characteristics to derive an average

packet size.

This over-simplistic design fails in two areas. First, it does not simulate a

realistic Internet environment, which is one of routing fluctuations, large

bursts of traffic and congestion. Second, packet-sized distribution is actually

multi-modal – meaning that IP traffic characteristics cannot be accurately

described by averages.

Another factor is the rate of traffic growth; enterprise traffic grows by 15

percent per year and can be accommodated by a legacy router through annual

upgrades and replacement after three years. However, IP traffic grows by 300

percent and, in order to accommodate such growth, the legacy router would need

an upgrade every month and replacement after six months. Purpose-built

architecture for service provision has to work well within its limits at initial

deployment, so that it can have a realistic usable life.

A Growing Need for Speed

Even with specially designed network architectures that go beyond the

"best effort" offering of legacy routers, problems, such as internal

fiber cuts and device failures, still occur. As the number of ISPs increases, so

does the number of interconnections. Since users and providers do not control

the architecture or operations of peer networks, these interconnections create

instability in the routing mesh. All these factors add up to a potentially

unstable network.

Some ISPs gain competitive advantage by carefully engineering their core

network traffic to maximize use of circuits as well as reliability. To

effectively engineer traffic, traffic rates and trends must be understood

between network entry and exit points. Also, there must be the ability to guide

traffic to available bandwidth, as opposed to simpler decisions such as the

shortest path through the network. ISPs formerly accomplished this by using

switches in their core networks and by creating switched paths that maximized

utilization of the network by routers.

Now, ISPs are exploring new technologies, specifically Multi Protocol Label

Switching (MPLS), to gain traffic-engineering capabilities on core routers. MPLS

speeds up network traffic by setting up specific paths diverted around congested

parts of the network–making the Internet quicker, easier to manage and more

reliable.

In this competitive market and potentially volatile networking environment,

service providers can no longer afford anything less than a reliable,

high-performance network designed and built for all conditions. Designing

networks based on scaled up enterprise equipment, around the principle of normal

operating conditions, is not sufficient and will not deliver IP traffic at the

speeds and quantities demanded in a growing broadband environment.

Value-added Broadband Services

The emerging broadband debate, together with competition and the growing

pressures for business advantage, are driving the need for multiple and diverse

service offerings, such as Voice over IP (VoIP), Video on Demand (VoD), Class of

Service (CoS) and Multicasting. In turn, bandwidth needs are increasing, as is

the scale of the Internet. During the phenomenal rise of the modern Internet, it

has, until very recently, almost exclusively carried data traffic. In turn,

voice services have always been carried in their own environments, on technology

that has been matured by incumbent telcos. Broadcast media such as television

and radio used to be carried over the air, cable or satellite.

Even video conferencing was developed to run over dedicated circuit links.

Today, all of these content sources are looking to Internet transport – as IP

allows diverse technologies to converge onto a single infrastructure. Using a

common IP infrastructure provides many advantages, not only for service

providers who can realize large operational cost savings by converging their

separated networks, but also businesses and consumers who are benefiting from

new services and applications. This broadband Internet is becoming an essential

business tool.

The only way for operators to cope with these new service requirements is by

implementing a strong and robust infrastructure, which can deliver data, video

and voice traffic at a quality of service appropriate for business critical

communications. As more and more diverse applications emerge, wire-rate

forwarding performance becomes even more imperative.

While it is acceptable for Web Pages to load in an irregular order, with

images following a few seconds after text, users of any real time video or voice

service, including VoD, would rapidly become disinterested if the multimedia

packets were not delivered in real time.

Building a Network for Scaleable Services

Value-added services require more than best-effort delivery; infrastructures

must transmit packets at wire rate even during congestion. For example, today’s

CoS requirements dictate that premium customers get priority access to bandwidth

during periods of congestion. With wire-rate performance, forwarding never

becomes a bottleneck and service providers maintain the ability to offer CoS

because there is no contention for processing cycles to execute both forwarding

and CoS queue management.

Service providers need to start future proofing their core networks now. As

broadband Internet usage evolves to include VoIP, interactive gaming

applications and other services which increase the volume of small packets, it

will become highly undesirable and expensive to change infrastructure. The costs

of adding nodes is significant enough but the hidden costs of designing,

testing, deploying and maintaining infrastructure can be overwhelming and can

even prevent service providers from growing at the rate they need, to keep up

with the competitive Internet market.

If service providers cannot adequately deploy the available bandwidth today,

how can they reliably offer enhanced services in the future? If a lightly loaded

system cannot guarantee wire-rate performance for all packets, under all

networking conditions, how can service providers grow it to a heavily loaded

system seamlessly, and without unnecessary upgrades?

New routing technologies are taking advantage of new hardware, software and

processing advances in the heart of a robust IP backbone to enable the broadband

Internet. This new generation of routers must maintain network control functions

and implement packet based services while achieving packet forwarding rates at

the highest speeds available from optical transmission technology. Furthermore,

these routers must perform these processes consistently, without ever blocking

or delaying traffic, so the network can live up to the promise of high speed,

multimedia broadband access.

Statistics show that broadband is on its way, and service providers are

starting to recognize that they must ensure their infrastructures are fast,

reliable and efficient. Only this will spell the end of the worldwide wait for

voice, video and data traffic. The broadband promise of real time multimedia

applications will not become a reality without the core deployment of robust IP

routers. It is the core technologies that will enable service providers to

deliver value added services, thus building revenues and recouping their vast

investments on upgrading access networks and ultimately securing their own

futures.

Alan Taylor, technical director EMEA, Juniper Networks

Characteristics and Benefits of Frame Relay