Peering into the future: peering and transit are the two most established methods to exchange internet traffic over different carrier networks. But another method is now emerging to challenge the old order��'advanced peering' - NGN performance and QoS - Industry Overview

Everyone wants to keep their costs down and internet-related companies are no exception. Due to cut-throat competition and pressure on prices and margins, considerable industry attention is focused on simplifying the logistics and reducing the cost of exchanging internet traffic between ISPs, cable operators, content providers and other interested parties.

The two best-established models for internet traffic exchange are peering and transit. Peering is settlement-free, where the traffic exchanged between the parties is deemed to be of equal value. Transit, on the other hand, is paid for--usually on the basis of the traffic volume carried or the capacity provided.

While peering has the advantage of being payment-free, it does have a number of disadvantages. "Negotiating agreements can be difficult with peering, particularly with the larger, more desirable peering partners. And don't forget the fact that peering often has to take place over public peering points which cannot offer service guarantees," observes Keith Mitchell, CTO of European metro area interconnect services specialist XchangePoint.

Herbert Goetsch, director of product management (global IP services) for international IP carrier, Level 3 Communications, argues that the actual cost structure that can be achieved with traditional peering is highly dependent on traffic volume and mix. Taken to the extreme, this type of peering can produce reverse economies of scale through the replication of resources and infrastructure. Goetsch also maintains that the participating networks operate at lower speeds than would otherwise be the case because traffic is spread across many networks. This has the effect of retarding the deployment of higher bandwidth technologies, such as Multi-Protocol Label Switching (MPLS).

Tony Marson, research director at Probe research, suggests that the contractual conditions imposed on the smaller players by the larger ones can be quite onerous, if not impractical, and there's the requirement to co-locate where the partners you want to do business with are. "The smaller players who cannot actually get peering agreements with the larger players will normally have to buy transit," he says.

Transit has a number of things in its favour. Mitchell points out that transit services are usually backed up by formal service-level agreements and that entry-level costs can be quite low. Marson suggests that the interconnect logistics are simplified, too.

On the downside, transit does cost money; and, the price barriers to entry may not be that Low where there is a dominant transit provider in a particular geographic market. Mitchell also reasons that the international transit routing obtained may not be optimal from the customer's standpoint.

It's the way that you do it

Options for peering and transit traffic exchange models include:

* paid peering, which involves fixed connection fees or recurring charges;

* settlement-based peering, where the sender of the largest volume of traffic bills the other party. ; and

* partial routing-based transit, when there is agreement to deliver traffic only to specified parts of the internet.

In partnership with select ISPs, cable operators and content providers, US-based Level 3 has developed a fourth option it calls 'advanced peering'. This is similar to the internet exchange model in which there are virtual links between peers over a shared ATM or ethernet switching fabric. In this case, however, the shared switching fabric is Level 3's MPLS network. In operation, each ISP, cable operator or content company connects to this network wherever convenient. Given the geographic reach of its inter-city and metro network infrastructure, Level 3 says that these connections will likely be fibre-based.

Once connected, each user determines who they want to peer with and enters into bilateral agreements in the same way as with traditional peering. Users then establish virtual links across the MPLS backbone and begin to exchange traffic. Multiple applications and traffic types--internet and non-internet--can be supported by this network.

According to Level 3, advanced peering delivers a number of benefits over traditional models. On the macro 'industry health' level, for example, it's much more efficient to concentrate traffic across fewer networks and achieve higher utilisation levels. These efficiencies, asserts the carrier, will act to accelerate the development and deployment of high bandwidth IP technologies, a phenomenon that in turn fosters further IP unit cost reduction and fuels the growth of additional high-bandwidth applications.

Not advanced pricing

Of perhaps more immediate consequence to ISPs and content providers--on the micro level--is that advanced peering is said to offer lower per transit costs than conventional peering. There are a number of components to this. One is a reduction in access infrastructure. Goetsch offers the example of European ISPs needing to peer with a US partner. "The advantage is they [European ISPs] don't need to set up a peering infrastructure network," he says. "They just need an access port in, say, Frankfurt or London or wherever, and Level 3 connects them to the access port of the peer, say, in San Jose. They don't need a port in San Jose, just a virtual connection to the network. "

Unlike traditional peering with self-provided access infrastructure, advanced peering users only pay for service they actually consume. "If you set up your own peering network infrastructure you need to fill it with traffic because right from day one you have fixed network costs. With advanced peering you just pay for what you use," states Goetsch.

Level 3 maintains that it is cheaper, for it to support advanced peering traffic rather than internet traffic. This is because internet traffic often traverses peering links, which means Level 3 will carry the cost for both origination and termination of the traffic, but will be compensated entirely by one customer. In advanced peering, Level 3 originates and terminates the traffic between two customers. These two customers in essence share the cost of the connection. Level 3's costs are approximately the same as if it delivered the traffic to a peering point, but the cost is shared between the two customers.

Level 3 says it is working with its customers to define a pricing model that reinforces the combined economies of advanced peering, internet transit and other traffic types. The resulting price per megabyte for advanced peering and other traffic should be 'materially' lower than transit.

In principle, running traffic over a single unified network minimises latency concerns that can arise when traffic is passed over a series of different interconnected networks; in general, according to Goetsch, the more hops there are in a 'multi-home' link, the greater potential for latency.

By the same token, though, electing to run over a single operator's network requires conviction that that operator is one step ahead of any availability, resilience and Quality of Service (QoS) issues. The network acquires greater consequence in light of the current industry expectation that prices will now be Low. "With Low prices taken for granted, the successful players are those who will manage to differentiate themselves from Low-end commodity offerings by supporting higher degrees of service quality as well as flexibility," predicts Mitchell.

Level 3 is keen to stress the MPLS base of its backbone network, which allows the company to offer QoS and Class of Service (CoS). Imagine an ISP wants to offer a premium gaming service for its, end users. An advanced peering solution enables the ISP to establish a direct virtual connection with a guaranteed QoS with the gaming provider even if that provider is in another continent. This kind of service is not currently sustainable on the public internet. The Level 3 network uses the Martini Draft implementation of MPLS, which defines how MPLS can be used to support Layer 2 VPN services, such as ethernet, frame relay and ATM over an MPLS-based network. It is designed to provide better management and control in running the MPLS network. Unlike similar protocols, it is does not require the deployment of routing tables for every customer on every switch and router throughout the network. The end result is lower operational costs.

Bust a glut?

Marson thinks it's probable in the future that the capacity glut will disappear on some routes. "We did an analysis of the US broadband market last year and we think that at the current rate of lighting they will run out of capacity in the next year to 18 months," he comments.

XchangePoint agrees. "In the longer term it is quite likely that the current capacity glut will exhaust itself, and we may even see supply/demand shortage/surplus cycles in the cost of transit as in other markets--for example, semiconductors," says Mitchell.

Lighting new fibre to meet any future capacity shortfall on particular routes-may not be viable for many tier 2 and 3 service providers if current constraints on capex continue. Goetsch believes that, as a consequence, the industry will witness the speeding up of an already discernible trend away from 'network build' to 'network buy'. "Two or three years ago most companies wanted to build their own network infrastructure," he says. "Today, everyone's really open to outsourcing network infrastructure, either by capacity or by service. "

In an unpredictable future, Mitchell predicts that success will go to the ISPs who are able to readily mix and match interconnect solutions. "In such an environment, the ability for an ISP to flexibly adjust its optimal balance of transit and peering without having to make major structural changes is a key advantage," he concludes.

Internet traffic still on the up

While it's possible to find analyses that warn of a slowdown in internet traffic growth rates, a 2003 report from the IDC consultancy--Worldwide Bandwidth End-User Forecast and Analysis, 2003-2007: More is Still Not Enough--predicts that the volume of global internet traffic will nearly double annually over the next five years, increasing from 180 petabits per day in 2002 to 5,175 petabits per day by the end of 2007.

"Some industry observers have speculated that slowing growth in internet traffic is at the root of the current telecom malaise, but IDC research shows that not only is internet traffic growth strong, but it will continue at near triple digit rates over the next five years," states Sterling Perrin, the company's senior research analyst for optical networks.

There are a number of catalysts at work here, quite apart from the on-going enlargement of the narrowband wireline internet user population and the boost that flat-rate internet access gives to traffic volumes.

First is the escalating uptake of broadband DSL access, cable modems and corporate metro ethernet services. A recent analysis from the UK-headquartered Ovum consultancy, for example, suggests that DSL deployments are currently beating earlier forecasts. At the beginning of 2002, the company forecast that there would be 35.3 million DSL lines installed worldwide by the end of that year. It now believes there were 36.3 million installed, or almost double the figure at the beginning of 2002.

"During 2003, we can expect rapid growth to continue, although not at exponential growth rates," says Ovum chief analyst, Julian Hewett. "We forecast 55 to 60 million DSL lines worldwide by the end of 2003. "

A second catalyst for the growth of internet traffic is the availability and commercialisation of new bandwidth-intensive internet-based applications such as corporate VPNs, gaming services and video streaming. And a third is the anticipated addition to traffic volumes that will be made as GPRS and 3G services gather momentum and technologies such as grid computing come on stream.

All this means that, notwithstanding the continuing industry turmoil, the market for internet services remains highly competitive and very price sensitive. According to Tony Marson, research director for the Probe Research consultancy, many ISPs are forced to continue to compete aggressively on price.

"The pressure on the ISP customer side is to reduce costs wherever possible. On the supplier side there's a glut of capacity so there's pressure to increase traffic as much as possible. " adds Herbert Goetsch, director of product management (global IP services) for international IP carrier Level 3 Communications. "As a result, you can see some incredibly Low prices in the marketplace. "

John Williamson, freelance writer

COPYRIGHT 2003 Horizon House Publications, Inc.
COPYRIGHT 2003 Gale Group