Articles

Feature Story       June 7, 1999

Surviving the Packet Revolution

Carriers looking to implement the access network of the 21st century must carefully consider how to maintain quality of service while running packetized voice and data over the same pipe

The industry has spent many R&D dollars in its search for new solutions for the core of the network, where more cost-effective and higher-capacity technology such as "big fast" routers, ultra dense wave division multiplexing and optical networking will ultimately hold center stage.

Once the dust settles, we will have arrived at a point where the entire communications traffic of a city the size of New York can be switched by a box that fits in the trunk of a mid-sized family automobile--and at a cost per switched bit that is two orders of magnitude less than the cost of a legacy circuit switch. Such heroics in the network core have recently been the source of much hype. But the core of the network is no longer where the action is.

The access network--the part of the network that delivers services to paying customers--has until now received little attention. This is changing as companies exchange the first salvos in the battle for the last mile. For service providers, this battle is about getting end users connected to their network and keeping them there. For end users, it's about getting better service at a lower overall cost. The access network is where the new heroes will emerge from, and it's worthy of a closer look. The battle is far from over in this arena, but we know that to survive the transition and emerge as a competitor, the successful 21st century network must have certain attributes.

It's no secret that data, specifically Internet protocol (IP)-based data, will be the dominant type of traffic on future networks. Accordingly, 21st century access networks will be data-centric. This new species of access network will be designed and built to carry packets as efficiently and effectively as possible. And all traffic (data and voice) will be packetized. This approach has a number of implications, not the least of which is how legacy services--and specifically voice--will be transformed to a packet format.

Advantage packets

Packet-based networks are a departure from traditional circuit-based networks in a number of ways.

First, packet networks have fully flexible access bandwidth. Legacy circuit-switched systems are at a significant disadvantage because of the rigidity imposed by channels or circuits and the inflexible multiplexing schemes that propagate these circuits through the network. The limitations of this access network, specifically in terms of its ability to deliver high-speed data services, include:

• Dedicated, inflexible bandwidth.
• Multiple infrastructures for multiple services.
• Difficult to add high-speed data support.

In the legacy world it was impossible to use the bandwidth released by an on-hook telephone to a bandwidth-hungry burst of data. Packetization of the voice and data streams at the customer premises allows these services to efficiently use all the available bandwidth. As a result, the 21st century access network will deliver fully flexible access bandwidth to end users, allowing multiple services--voice and data--to share the access pipe.

Another advantage of packetized networks is that they have a single access infrastructure for multiple services. Service providers today have multiple infrastructures delivering different services, which is a result of how their networks were deployed. The voice network has been around the longest and was followed by a number of overlay networks, each with its own mission such as DDS or frame relay. Each service is individually backhauled from the customer's local point-of-presence to the central office, each on its individual backhaul facility.

But multiple infrastructures mean increased cost in terms of capital and ongoing maintenance. Again, reducing all traffic to a single flexible packet format allows different traffic types to share the same backhaul facilities, thereby providing a single infrastructure that supports multiple services. The ability to be service-independent is critical because carriers don't know what services their customers will want to buy.

In addition to presenting a fully flexible, multiservice-capable customer access line, the networks of the future will be fully managed. Operations support systems (OSSs) are the true enablers of delivering high-quality customer support. The ability to accept customer orders, deploy service, maintain the quality of service (QOS) at or above the contracted level and accurately bill the customer are critical OSS functions.

The packetized future also will have "flatter networks." Today's legacy networks were built within the constraints of a technology hierarchy well suited to a regulated environment. With the pace of deregulation accelerating worldwide, such a hierarchical approach is no longer necessary, facilitating the possibility of distributing functionality throughout the network. In addition, building such hierarchical networks today is cost-prohibitive. The 21st century access network will comprise a collection of flexible, multi-mission, intelligent network elements that are distributed throughout the network.

In the same way as legacy voice networks today serve tens of millions of customers with voice services, 21st century access networks will deliver high-speed data as well as voice to an even larger population. In addition, the packetized networks will equally serve major metropolitan areas and small villages, business parks and multitenant buildings.

A cost-effective and scalable switching fabric--along with more powerful digital signal processing capabilities--will make it possible to achieve this goal. This represents significant incremental opportunity for service providers, which previously could not provision the small customer at a rate that was competitive with the large, corporate customer. The competitive environment will change considerably as new, smaller service providers are empowered to compete with more established, legacy network providers. The end user will have greater choice of providers, making issues such as QOS and overall customer service important differentiations.

Issues to consider

Carrying data on a packet network is well understood. So is the concept of offering different grades of service for different applications, such as giving a higher priority to--and charging more for--mission-critical data. Carrying voice over IP on a packet network should also be familiar for carriers because IP telephony is simply another data application. This is not the case for traditional analog voice.

A bigger challenge is carrying traditional analog voice in a packet format integrated with data on the same access line. This is important because traditional voice is going to be around for some time. All the world's telephones are not going to be replaced overnight by voice-over-IP terminals. And even though voice will, over time, be replaced by data as the primary source of revenue, traditional voice will remain a critically important element of the service provider's portfolio.

Because data replaces voice as the profit king, it is critically important to carry voice in the most cost-effective fashion (on the same packet infrastructure that carries data) and with the same quality as modern digital systems provide today. This does not mean, however, that data-only service providers can relax. On the contrary, because customers will require network operators to deliver a complete service portfolio, those that can't will constantly face the threat of being displaced by a competitor that can.

Handling voice places some specific requirements on a packet-based access network. Although voice requires little bandwidth, it has strict latency requirements. In a mixed voice and data packet network, the key challenge is to ensure low latency for voice while handling data packets as efficiently as possible. In other words, carriers shouldn't penalize the data by carrying voice or degrade voice quality by carrying data. A number of approaches can be adopted to optimize this balance:

• Packet prioritization.A service provider can assign a priority to an individual packet by manipulating a label in the packet header. Higher-priority packets gain the right to "bump" lower priority packets in order to get to the top of the queue. Significant work is underway to bring guaranteed QOS to packet networks, specific schemes include multiprotocol label switching. Applying the highest priority to voice packets ensures that these packets receive the highest priority from the network element resources.
• Latency management. Larger packets take more time to process onto physical links than smaller packets. No matter what size it is, once a packet has been passed from the queue to the link, it's gone. If this were a very large packet, then the following packet--no matter what its priority--would have to wait until the long packet is processed. This can be a problem particularly on slow links. It is therefore necessary to limit the maximum packet size in order to manage the latency of the system. Network operators should manage all traffic queues in such a way as to ensure a maximum bounded delay for the highest priority packets.
• Voice quality assurance. The main issue related to voice quality is echo control. Because packetizing voice (for example, collecting encoded samples into a packet) takes some time, this delay, when added to other sources of delay in the network, necessitates the use of echo cancellation in order to provide high-quality voice. Such systems are used extensively today in digital cellular networks for precisely the same reason.

A number of key issues need to be fully understood by network operators that are building packet-based access networks. Building a network that is reliable and that offers a range of quality services is a significant first step on this road to a 21st century access network. If an emerging competitive carrier can build a packet-based access network with fully flexible access bandwidth, sophisticated OSSs and a scalable, "flatter" topology while offering packet prioritization, latency management and voice quality assurance, it will be counted among the survivors when the smoke clears on the 21st century access landscape.

Top of Page