Broadband access: the wave of the future.

Technology adjusts to packets and must offer QoS.

You have heard circuit switching is dead. The wave of the future is packet technology. Yet, sales of class 5 switches show no sign of slowing. Still, sophisticated new Internet applications such as e-commerce, application hosting, streaming audio and video, and tiered Internet access, are driving a fundamental restructuring of the telecom infrastructure into one based on packet technology. Understanding the transformation to packet can help service providers and enterprises avoid costly mistakes and build future-proof networks.

The beginnings of the revolution were first felt at the core of the public network with the deployment of packet-over-SONET (POS) and dense wave division multiplexers (DWDM). The waves were then felt in the next layer of aggregation in the public network with the deployment of terabit routers as an aggregation device. The waves have now hit the local loop, the much-discussed "last mile."

Because of its speed, reach and security, digital subscriber line (DSL) is preferred for local loop. To fulfill its potential, however, DSL needs to evolve in two fundamental ways:

* It needs to adjust to the packet revolution in the public switched telephone network (PTSN) that these new services have precipitated; and

* it needs to provide the quality of service (QoS) required by the broad range of applications end users are demanding.

Today's generation of DSL access multiplexers (DSLAMs), while more flexible than first generation DSLAMs in that they are designed to enable multiple services (Internet, voice, frame relay, virtual private networks), still only segregates and aggregates voice and data streams. DSLAM is simply an access multiplexer. In order to support the more sophisticated services, the next generation of DSLAM needs to address the requirements of these applications by intelligently classifying the access streams and prioritizing the level of service the applications require.

The shortcomings of the current-/second-generation DSLAM that the next generation needs to address include:

* Lack of scalability associated with virtual circuit-oriented service delivery; end-to-end handling of QoS requirements for each application across multiple service provider networks and the public Internet; and

* differentiation between the various application traffic streams--voice, intranet/extranet access, e-commerce, Web hosting, application hosting, videoconferencing, streaming audio/video--in an Internet access pipe.

SCALABILITY CHALLENGE

Foremost among the problems facing DSL operators today is the issue of scalability. The single largest obstacle to growth for a DSL carrier is the complexity of the prevailing Layer 2 model that creates a permanent virtual circuit (PVC) mesh that expands on an exponential basis.

Using the Layer 2 provisioning model, each subscriber service (e.g., voice or data) per user requires an individual PVC to be established on an end-to-end basis. In a simple example with 200 subscribers in a central office (CO), 400 virtual circuits (VCs) are created. As the carrier expands market footprint to include more COs in the region, tens of thousands of VCs are added; 100 COs results in an astounding 40,000 PVCs.

This degree of complexity presents several practical problems;

* High operations overhead. Large VC meshes entail huge costs of operation. Manual PVC provisioning wastes thousands of hours of provisioning time; it also slows the process of turning up each new customer.

* Difficulty of maintenance. A 40,000-PVC network is clearly unmanageable. Network operations staff would be quickly overwhelmed if asked to identify and troubleshoot an individual flow problem in such a complex network.

* Platform limitations. Even carrier-class asynchronous transfer mode (ATM) switches would be hard-pressed to support a 40,000-PVC network, and today's routers and the cost for additional routers, voice gateways and service-management systems (SMS) simply for VC aggregation would be astronomical.

Although "band-aid" Layer 2 solutions have been proposed, none have proven to be practical. For example, switched virtual circuits (SVCs) have been suggested by a few DSLAM vendors, but today no DSL integrated access device (IAD) or router supports this capability. Similarly, mainstream voice-over-DSL gateways lack support for switched virtual circuit (SVC), as well. Finally, SVC performance of most carrier-grade ATM switches, at best, is lackluster. The answer to this critical issue is to leverage the inherent scalability of Internet protocol (IP) at Layer 3, instead of simplistic Layer 2 mechanisms.

The second issue facing DSL providers is implementing a practical end-to-end QoS scheme, tightly linked with the immediate needs of the application. Ideally, QoS is "signaled" on demand by the application. Although Layer 2 ATM and frame relay standard QoS schemes exist, PVCs are effectively "preconfigured" by priority and cannot react to dynamic changes in application requirements or network resources. A growing number of applications now utilize IP QoS standards-based mechanisms in order to specify the QoS requirement of each individual packet.

SIMPLE, SMART NETWORK

Since early this year, a growing number of DSL equipment vendors have been publicly disclosing solutions based upon Layer 3 (IP) technology. Multiprotocol Internet scaling architecture (ISA), for example, enables DSL providers to scale their access networks to large proportions, while enabling new differentiated access services using standards-based IP mechanisms. The three core elements of ISA are: IP services aggregation (IPSA), multiprotocol label switching (MPLS) and IP quality of service.

IPSA provides a new model for DSL providers to rapidly enable high network scalability, while continuing to leverage their investment in existing ATM switches, routers and gateways. IPSA solves the scalability problem by using IP-level intelligence to merge IP flows in the upstream direction to the point of presence (POP), and to forward packets to the correct port and VCs in the downstream direction to the customer-provided equipment (CPE).

Multiple IP forwarding classes can be established to create logical groups of IP flows that share a common IP route and require the same class of QoS. Referring to the PVC example demonstrated earlier, the total number of PVCs in a market of 100 COs drops to only 200 PVCs total.

MPLS is rapidly becoming the de facto standard for end-to-end IP flow provisioning within ISP backbone networks. In effect, MPLS marries the intelligence of IP routing with the high performance and simplicity of Layer 2 switching. Several factors are driving the MPLS revolution in carrier-class ISP/ICP networks:

* High scalability for IP provisioning over existing Layer 2 technologies (ATM, frame relay, and point-to-point protocol);

* multiple IP-based classes of service for differential path selection based on the QoS of each flow;

* reduction in the overall complexity and cost of network operation by enabling traffic engineering; and

* standards-based mechanism for multilayer switching.

In their roles at the network edge, intelligent DSLAMs add significant value by complementing high performance MPLS label switch routers that are projected to replace older-generation ATM switches in the POP. While these new gigabit and terabit-class devices will provide tremendous performance and bandwidth, they still require MPLS labels to be initially inserted in the packet flow, and later stripped prior to handing off to a non-MPLS Layer 3 device.

A primary enhancement under the Internet scaling architecture is support for standards-based IP QoS mechanisms. Most IP-native networking applications requiring better than best-effort services will utilize one of the following mechanisms supported under the ISA model:

* IP type of service (TOS). The TOS field consists of three bits in the IPv4 header. Intended for use by applications requiring a simple scheme to signify a given packet's relative priority, IP TOS is in use today by several applications, including voice over IP-based IADs and voice gateways. Intelligent DSLAMs will support these applications by classifying each packet in the flow based on TOS value. A user-configurable mapping will resolve to which priority queue the individual packets will be linked for output scheduling.

* MPLS class of service (COS). The COS field is three bits in length, and follows immediately after the 20-bit MPLS label in the MPLS header. COS is used as a mechanism to help queuing and discard decisions at a packet-level as the packet transits the MPLS label-switched path. PacketLoop i-SLAMs will support the COS field for assignment of packet flows to specific forwarding equivalence classes and for output buffer queue scheduling.

* IP differentiated services (DiffServ). DiffServ is a standards-track framework for a simplified differentiated service delivery provided on a per-hop basis. DiffServ allows performance requirements to be specified based on a traffic conformance specification (TCS). A TCS may be established that specifies traffic parameters such as sustained bit rate and burst size.

So, while circuit-switching is far from dead, the entire public network--from the core to the edge--is being gradually restructured, with packet technology forming the building blocks, to accommodate the explosive demand for bandwidth and higher-speed Internet access.

www.accesslan.com Circle 262 for more information from AccessLan Communications, Inc.

[ILLUSTRATIONS OMITTED]