DSL and QoS: walking hand in hand; Quality of service for DSL will be mandatory for enabling advanced services—especially video

Telecommunications Americas, Nov, 2004 by Bert Whyte

There's a lot of interest in roadmaps for evolution of carrier infrastructure, and as the number of deployed DSL lines rapidly approaches 100 million, one focal point is the inevitability of running more advanced services over the existing DSL network. The economic case for re-use is clear, since the alternative is to fund a partial or wholesale replacement of access infrastructure, but can a compelling technical case also be made?

QoS and reliable end-to-end policy enforcement are key characteristics of many applications. With IP established as a cornerstone of service strategies, the need and mechanisms for implementing these features in non-IP-aware DSL equipment has generated a lot of discussion.

[FIGURE 1 OMITTED]

The first question is whether telcos need QoS in their DSL networks. To put DSL QoS in context, it helps to start by looking at the bigger picture for carriers, who, whatever their size, are being driven to re-define and re-vitalize their business models. And it's not only because traditional voice revenues have been eroded by price competition in an apples-for-apples market. A more worrying threat now exists from cable MSOs, who have upped the stakes by adding voice services to create fully integrated triple-play packages. Triple play used to be an interesting possibility for carriers, but now the lack of it is a competitive disadvantage that must be addressed.

Suddenly, we're into the realms of mandatory DSL QoS, for the simple reason that without it any attempt to deliver video services will come unstuck.

To see why, consider a DSL line carrying a combination of video data (30 frames of video per second ... bang, bang, bang ... with little room for variable network delay) and inherently bursty Internet traffic (text, plus images, plus file downloads). It is surprisingly easy for the Internet data to upset the smooth flow of video and cause visible damage to the picture. The only defenses are to either provide significant overcapacity on the link (increasingly unlikely for DSL, as service demands increase) or to prioritize traffic to ensure the video gets through in a reliable and timely manner and is not adversely affected by the best-effort web traffic.

In a more general sense, with packet voice, gaming and other services joining video in requiring priority access to the DSL link, the need for flexible DSL QoS becomes acute. The speed of the DSL connection is increasingly unimportant when considering the need for QoS--VoIP services using today's midband DSL connections are becoming more common and require protection from simultaneous web browsing sessions. But it's the need for video that has galvanized acceptance of DSL QoS as a generic, mandatory requirement.

To be successful, any enhanced service must meet or exceed the triple play offer from the competition. In addition to the voice and data elements, for some carriers this means support for two standard definition TVs, while for others three TVs is the minimum requirement, including at least one high-definition set. The resulting bit rate could be anything from 4 Mbps to 20 Mbps or more. The network demands of video may allow re-use of existing ADSL, or may be beyond the capabilities of the existing access technology and require equipment and/or network upgrades to accommodate ADSL2/2 , VDSL, VDSL2 or FTTH.

It comes as a surprise to many that the choice of access technology may have little effect on the need for QoS. It may be possible to live without it for a while in cases of significant overcapacity, but it's expected that growing service demands will mandate QoS even in situations where very high DSL bit rates (by today's residential standards) are provided. But it's the need for video which has galvanised acceptance of DSL QOS as a generic, mandatory requirement (See Figure 1).

At this point, it's worth mentioning that QoS in the new world order is not simply about getting bits reliably from A to B. It's also about reacting to a constantly changing service profile as demands appear, change and disappear. Supporting controlled prioritization between any number of service providers on the one hand and any number of customers on the other, and all without the need for human intervention, is critical. Effective QoS requires dynamic flexibility on a moment-by-moment, end-to-end service basis.

IP QoS in the Access Network?

So DSL QoS must be present, must be flexible, and is largely independent of both bit rate and access technology ... but how should it be implemented? To help make an educated and objective decision, we should briefly look at infrastructure trends in the broader network.

There's no doubt that carriers are increasingly building their service strategies around IP. In the core this is reflected by procurement migration away from ATM and toward Ethernet, and by the intense interest in MPLS as a way of enabling existing ATM fabric to more easily and readily accommodate IP traffic.

[FIGURE 2 OMITTED]

As the move toward IP progresses, IP QoS coupled with end-to-end policy enforcement will be used to prioritize complex and continuously changing traffic profiles through the core and into the regional networks.