Virtualizing DSL delivery: providers need to make the leap from broadband conduit to service delivery

Broadband access using DSL technology is still in its infancy and going through a series of growing pains. From its inception, DSL has been widely viewed as a way to satiate the needs of Internet power users. These end users have no problem finding new ways to utilize their broadband connection and, more important, justify the price.

But as service providers look to penetrate the second and third tiers of customers--those who primarily use the Internet for e-mail and other nonbandwidth--intensive applications--they must provide more compelling reasons for customers to upgrade their dial-up connections. Displaying the power of broadband connectivity through applications such as telephony, multimedia messaging and video on demand becomes paramount. To drive the most value and attract the most customers, service providers must begin to transform the way they think about broadband. Instead of being a simple conduit, DSL must be considered a service delivery vehicle.

Along with this philosophical change, service providers must contend with the realities of today's networks. Challenges include:

* Provisioning. Today, and for the foreseeable future, most DSL services are delivered using ATM. However, ATM service in the edge and core causes scaling issues, since one or more PVCs are dedicated per subscriber. In addition, ATM does not make it easy for service providers to troubleshoot and manage their networks since it is difficult to attain an end-point-to-core view of the network. In traditional ATM-based DSL architectures, the DSLAM is not visible within the troubleshooting path, making it difficult to identify whether problems exist in the CPE or the DSLAM. Moreover, services that could be optimally routed in IP are carried back to a central point before they are routed, adding unnecessary layers, decreasing scale and increasing cost.

* Extending services to the entire subscriber base. Many customers cannot receive DSL service because they sit behind analog remote terminals or are simply too far from the serving area to make it worthwhile. Over time, this problem will be addressed through upgrades to existing remote equipment and the deployment of more intelligent, broadband-enabled switching and access equipment. However, service providers must choose their customers wisely; those who are farther from the central serving area tend to be more elastic in their demands for advanced services.

* Delivering services cost-effectively. Today's DSL delivery networks exist as overlay networks, adjuncts to the traditional voice services infrastructure. They are costly to maintain and upgrade, and thus far have produced unsatisfactory returns on investment. Combining the two networks and addressing both traditional analog subscribers and broadband subscribers is one of the ways service providers can deliver broadband services more cost-effectively.

Service providers can move upstream from basic connectivity and bandwidth to services by moving to IP/MPLS-based service delivery platforms. Moving to IP allows service providers to more effectively deliver a wider range of features and services, particularly broadcast video services based on multicast. Moving to MPLS allows service providers to provide secure, MPLS-based VPN services to subscribers for telecommuting applications. In addition, third-party retailers can gain visibility to end users through MPLS-based VPNs and provide complementary voice and data services, thereby virtualizing DSL delivery.

Virtualizing the DSL Infrastructure

In order to cost-effectively transform their existing DSL network infrastructures from bandwidth-delivery vehicles to service-delivery vehicles, service providers are looking for solutions that achieve these objectives:

* Unified voice, data and video. Continuing broadband overlay network deployment without unifying services ensures that the operational and maintenance costs will be higher than a comprehensive, unified infrastructure.

* Service quality. When running multiple services over a unified infrastructure, QoS must be maintained so that performance liabilities do not occur.

* Cost effectiveness. A unified solution must be cost-justifiable for every service. The unified infrastructure should leverage installed investments where it makes sense, and cost-effectively add new components to complete the service delivery transformation.

* Scalability. A unified network must be capable of growing over time to accommodate an increasing base of subscribers, as well as higher bandwidth.

* Standards-based. A standards-based implementation guarantees service interoperability, quicker time to market and ROI, better management of services, and easier back-office integration among various network elements.

* Flexibility. The unified infrastructure must also offer an intuitive service creation environment for multiple broadband services to guarantee service differentiation, innovation and competitive advantage. With the ability to build new and unique content and services for the end user, service providers can maintain customer loyalty and subscriber base growth. In addition, service providers can offer subscriber self-provisioning and personalization of all features and services.

All these mandates can be achieved using an IP/MPLS-based architecture. In addition, by using MPLS-based VPNs, service providers can enhance DSL service delivery by customizing the level of control and types of unique service offerings they can provide to subscribers. For example, subscriber flows, regardless of type, origination or termination, can potentially be mapped into unique VPNs and metered as an aggregate system. The service provider can then define and map user flows to VPNs based on subscriber or service type, thereby reducing the complexity of managing a unified broadband network (see Figure 1).

In addition, using MPLS-based VPNs for broadband service delivery, service providers can also:

* Expand the multicast address space. Multicast relies on Class D addressing, and is limited in the number of multicast streams that may run concurrently over the network. Using MPLS-based VPNs, the service provider can create VPN user groups and scale the number of multicast streams by multiples of provisioned VPN user groups.

* Meter services based on type. MPLS-based VPNs per service type may be created for simplified bandwidth metering. For example, the service provider can create three primary service VPNs-one for voice, one for data, and one for video. The video VPN will require the highest amount of bandwidth and is the least tolerant of jitter. The service provider can use the provisioned VPNs and their associated threshold limit to meter the service flows operating over each VPN.

* Wholesale services. MPLS-based VPNs can be used to wholesale services to third-party service providers. Services can map to specific ISP VPNs based on subscriber and type of service, and then be handed off to the appropriate service provider.

* Provide QoS. Services such as video, delivered using IP multicast, are sensitive to network latency and packet loss. IP can offer a level of control greater than class of service (CoS) within a routing domain, but becomes cumbersome and ineffective when crossing multiple IP domains with differing bandwidth characteristics. The simplest solution is to build a well-defined, multicast-capable network over which multiple services will effectively operate. Using IP+MPLS flow metering, service providers can establish an effective system of addressing QoS Figure 2 shows that a physical connection can partition into 256 unique private networks, each with its own class A-D address space. Each defined VPN will have a maximum and metered allotment of bandwidth. The total bandwidth of all 256 VPNs never exceeds the physical layer's maximum data rate. Within each of the 256 private networks, 64,000 sub-ports can be defined. Typically, one sub-port is assigned to a subscriber. That sub-port has a maximum allowable bandwi dth that is used to constrain the subscriber's maximum throughput in the event of contention. Within each subscriber's pipe, eight levels of GoS can be defined, with voice, data and video being assigned different types of service values. Finally, each ToS is queued within its own weighted fair queue.

Re-casting DSL as a service delivery vehicle utilizing an IP/MPLS-based infrastructure is one option service providers have for increasing their subscriber base and competitive differentiation in the nascent broadband services industry. The IP/MPLS combination affords service providers an increasingly larger array of features and services. Service providers can also enhance this baseline functionality by using MPLS-based VPNs to increase the overall efficiency of their DSL service delivery networks. Providers can aggregate subscribers based on type of service in the access network, as opposed to the edge or core, thereby reducing the number of layers required to successfully manage DSL subscribers. In addition, they can provide customized and personalized features and services to their subscribers to ensure they will find value in broadband beyond its usual positioning as a simple conduit.

William Flanagan is vice president of marketing for sentitO Networks (wflanagan@sentito.com).

COPYRIGHT 2002 Horizon House Publications, Inc.
COPYRIGHT 2003 Gale Group