The cost economies of Ethernet, coupled with the tremendous investment carriers have already made in Ethernet infrastructure, continues to drive Ethernet as the technology of choice in Metro applications. Year after year Ethernet persists in overcoming whatever new challenges the growth and convergence of networks brings by evolving through new innovations and standards.

However, as carriers begin to deploy "Triple-Play" networks carrying voice, video, and data traffic for residential markets and converged data services and hosted VoIP in the business enterprise network, the real-time demands and quality of service (QoS) requirements of these latency-sensitive applications have raised the bar for Metro Networks beyond simple best-effort service. Ethernet must once again adapt itself to provide the scalability and performance required to transport real-time traffic with carrier-class reliability.

Multidimensional Ethernet

It is important for both engineers and service providers to understand the subtle differences between residential and business triple-play services. Consider the impact of real-time video on the network. While residential customers are primarily interested in receiving video, in general, businesses have little to no interest in IPTV. Rather, their focus is on video broadcast for applications such as training, customer service, and teleconferencing. Voice services similarly confuse network topologies as networks move from circuit-switched to IP-based VoIP.

Given the goal of Metro Ethernet to make access less expensive, carriers ideally would prefer a single piece of equipment that services both business and residential customers rather than segregating them through different equipment as is often done now. Additionally, QoS is no longer an issue isolated to the local network. As the network becomes increasingly packet based, QoS must be pushed out to the edge and on to WAN links as well. To implement this, Metro networks will need to implement Multidimensional Ethernet.

Multidimensional Ethernet extends the reliability, scalability, and performance of Ethernet through four critical technologies: Hierarchical Quality of Service (QoS), Ethernet Cross-Connect flexibility, MAC-in-MAC scaling, and carrier-class Service Resiliency. By utilizing Multidimensional Ethernet, service providers will be able to implement both residential triple-play and converged business networks on the same hardware platform while increasing capacity, improving quality of service, and reducing overall operating expenses.

Hierarchical Quality of Service

The convergence of business and residential services requires an increase in the granularity of quality of service that can be provided by carriers. In the past, quality of service has been focused on providing enough classes of service to differentiate between data and latency-sensitive traffic such as voice. Carriers, however, have learned that they can increase overall revenues by differentiating between customers as well by offering multiple levels of bandwidth and guaranteed service. Additionally, with the increased number of customers being served, it becomes increasingly important to be able to provide deep QoS support over high-speed connections carrying tens of thousands of flows.

Hierarchical QoS utilizes hardware-based controllers to manage bandwidth and priority on a per subscriber basis while supporting multiple QoS levels per subscriber (see Figure 1). In this way, service level agreements (SLAs) can be guaranteed all the way down to individual flows. Put in concrete terms, Hierarchical QoS enables scaling that is 40 times larger compared to traditional VLAN implementations.

Figure 1. Hierarchical QoS utilizes hardware-based controllers to manage bandwidth and priority on a per subscriber basis all the way down to individual applications. Hierarchical QoS offers 40 times the scalability of traditional implementations.

Ethernet Cross-Connect Flexibility

In both residential and business applications, there is competition to provide content to customers. However, no one wants multiple pipes running to the home or business when a single connection can suffice. For example, businesses are seeking to move away from leased lines to hosted or internal VoIP implementations to significantly reduce overhead. Traditionally, carriers serviced each market with independent networks; today they need to be able to do it all through a single converged system.

The challenge of efficiently provisioning services is not trivial. Residential customers have a choice of television and voice service providers, as well as ISPs. Businesses not only utilize multiple content providers but face the additional difficulty of needing to connect multiple sites that are both logically and geographically separated. On top of this, both types of subscribers continue to demand more services and more bandwidth over existing connections. Finally, provisioning between multiple content providers needs to be flexible as well as simple in order to accommodate subscriber choice without undue burdening and network management.

Ethernet Cross-Connect functionality provides an efficient means for switching individual subscribers to different -- and multiple -- content providers (see Figure 2). By handling service connections at layer 2, service providers are now able to truly serve both business and residential customers from the same box. Convergence of business and residential traffic yields tremendous benefits and return on investment. For example, the consolidation of infrastructure equipment reduces network management complexity and personnel requirements which can lower deployment costs significantly. In addition to the savings in operational expenditures, using a layer 2 Ethernet cross-connect provides up to an 8X cost improvement for the network hardware over a typical layer 3 implementation.

Figure 2: Ethernet Cross-Connects provide an efficient means for switching individual subscribers to multiple content providers by handling service connections at layer 2. Convergence of business and residential traffic enables consolidation of infrastructure equipment, reduction in network management complexity and personnel requirements, and can lower deployment costs by up to 8X compared to a layer 3 approach.

MAC-in-MAC Scaling

Scaling is one of the most critical issues in Metro networks. The sheer number of flows an individual switch must accommodate in the Metro can undermine the performance and flexibility of existing tunneling technologies. Over the years, efforts have been made to increase the robustness and efficiency of Ethernet through technologies such as Layer 3 MPLS which leverages the speed of Layer 2 switching into Layer 3 by cleanly separating forwarding control functions from payload data. In the case of Layer 3 MPLS, however, overburdened core routers have to maintain a separate Virtual Route Forwarding (VRF) table for each VPN, exposing private routes associated with each subscriber and forcing a compromise of control and security of these networks. Layer 2 MPLS attempts to resolve these shortcomings but introduces new ones, including limited scalability and inefficient handling of multicast traffic.

The proposed MAC-in-MAC standard, formally known as IEEE 802.1ah, builds upon existing VMAN technology (see Figure 3). Service providers can manage individual subscriber traffic while supporting multiple levels of QoS by using VMANs to tunnel through VLANs. MAC-in-MAC implementations overcome the inherent scalability limitations of VMANs -- VMANs are bound by the limitation of 4096 VLAN IDs, making them impractical for use in large Metro networks and provider backbones -- by adding a transparent provider MAC address to the Ethernet frame. Doing so enables 4000 times as many service VLANs as supported by traditional VLAN and VMAN networks, effectively enabling service providers to economically scale to millions of service VLANs. MAC in MAC also neatly avoids Layer 2 scaling issues by eliminating the need for core and backbone switches to learn hundreds of thousands of MAC addresses. As a result, MAC-in-MAC provides carrier class scalability and reliability while reducing overall network complexity. It also enables seamless interoperability with existing VMANs without introducing the complexity of MPLS.

Figure 3: The proposed IEEE 802.1ah (MAC in MAC) standard works in conjunction with vMAN (Q in Q) technology. A provider MAC address is added to the Q in Q tagged Ethernet frame, effectively enabling service providers to economically scale to millions of service VLANs while eliminating the need for core and backbone switches to learn hundreds of thousands of MAC addresses.

Completing the Reliability Ring

While deep QoS, simple provisioning, and scalability form the foundation of residential and business triple-play networks, it is critical that engineers and service providers also address the reliability of Ethernet networks. Circuit-switched networks, for example, provide subscribers with telephony services backed by carrier-class "five nines" reliability. In order for Ethernet to supplant existing circuit-switched networks, it must not only offer a wider range of services but match or exceed such reliability as well. Additionally, failure avoidance and recovery mechanisms must extend as far as the applications and services supported by the network.

This concept is known as Service Resiliency, and it is an extension of Network Resiliency. Network Resiliency refers to high-availability and redundancy features such as redundant management modules, backup power supplies, and flexible switching fabrics which work in conjunction to provide protection against most device failures and offer hitless switchover from primary to backup units in the event of equipment failure. Network Resiliency, however, while guaranteeing that a broken connection will be quickly restored, does not guarantee that the myriad of services carried over this connection will not be interrupted.

Ethernet Automatic Protection Switching (EAPS) -- also known as IEEE RFC 3619 -- is a widely installed protection mechanism for native Ethernet interfaces. EAPS is a service-aware protocol that utilizes a standard Ethernet MAC and a variety of ring topologies to provide carrier-class failover response within 50 ms.

EAPS also enables service providers to select primary/backup designations on a per-VLAN basis. Such configurability down to the VLAN level promotes route diversity and provisioning of service flows across multiple switches and rings to maximize spatial reuse while minimizing the possibility of a service outage triggered by failures in other parts of the network. When working in conjunction with technologies such as the proposed IEEE 802.1ag CFM (Connectivity Fault Management), protection mechanisms such as EAPS harden Multidimensional Ethernet networks, offering both reliability and manageability in a cost effective Ethernet platform.

Multidimensional Ethernet provides the critical elements for bringing carrier-class performance and reliability to the Metro network. Coupled with Service Resiliency mechanisms, these up and coming next-generation Ethernet networks are poised to dominate the Metro as they have the Enterprise. By introducing innovations such as Hierarchical Quality of Service (QoS), Ethernet Cross-Connect flexibility, MAC-in-MAC scaling, and carrier-class Service Resiliency, network operators will be able to converge residential and business services to improve network visibility, increase throughput capacity, and reduce total cost of ownership while simplifying overall network topology.

Harpreet Chadha is senior product manager for Extreme Networks.

Visit Extreme Networks online.