For years, Ethernet supporters have predicted that Ethernet will become the dominant network technology and lead to a truly homogeneous network. Even as new standards have threatened to provide functionality beyond the scope of Ethernet, relentless innovation has led to the evolution of an Ethernet that can provide unparalleled scalability, reliability and Quality of Service (QoS) for next-generation Metro networks. The result, Multidimensional Ethernet, is a combination of MAC-in-MAC, Hierarchical QoS, Ethernet Cross-Connect, and Service Resiliency technologies that maintain Ethernet's ability to meet the specific functional needs of the rapidly growing business services market. It is through Multidimensional Ethernet that Carriers will be able to deploy truly converged networks over an economical network infrastructure while providing QoS to support the real time data requirements of voice, video and data applications.

Virtual Private Network Background 

The foundation behind enabling business services is the Virtual Private Network (VPN). VPNs enable carriers to break free of the physical restraints of a particular network infrastructure by logically separating services from the physical network topology. This is achieved by mapping logical groups of separate networks to specific tunnels, making it possible to connect, for example, offices that are geographically distinct as if they were part of the same LAN. These tunnels encapsulate traffic, providing a secure mechanism for transporting traffic between different locations as it traverses all the intermediate networks between them (see Figure 1).

VPNs have revolutionized networking by enabling remote networks to transport traffic over networks based on different physical infrastructure and protocol standards than the originating network. As new and innovative application spaces have arisen, VPN standards have evolved as well, extending the functionality of networks and the options available to carriers. 

The basis of scaling Ethernet networks is the IEEE 802.1Q Virtual Local Area Network (VLAN) standard which enables carriers to isolate and manage groupings of traffic as independent entities without regard for physical network connections. VLANs, however, were designed for simple Enterprise LAN environments, not Metro networks. As a consequence, there are inherent scalability limitations, such as the fact that VLAN IDs are only 12 bits; it doesn't take long to exceed 4096 VLANs in Metro applications.

The IEEE 802.1ad Q-in-Q standard extends the intelligence and scalability of VLANs by enabling carriers to support a service VLAN tag. This allows the carrier to grow the number of possible service instances, with a typical deployment consisting of a Service VLAN per subscriber. While this does not make the most efficient use of the combined Customer-VLAN and Service-VLAN address space, it allows the customer to maintain the VLAN structure they have in place on their local area network when they subscribe to a Carrier Ethernet service.

Q-in-Q tunnels are completely isolated from other tunnels, providing private point-to-point or point-to-multipoint connectivity across the network. However, while they provide a substantial improvement in scalability, they still lack the capacity to serve very large regional networks and provider backbones because of the persistent limitation of the VLAN ID size. 

Originally developed to improve performance by cleanly separating forwarding functions from the payload data, Multi Protocol Label Switching (MPLS) is often used to deliver Layer 3 VPN services to business subscribers. Layer 3 MPLS VPNs, however, require private routes associated with each subscriber to be shared with Service Providers, creating the undesirable result that subscribers must relinquish some control and security of their network. These private routes must be stored in a Virtual Route Forwarding (VRF) table per subscriber on the Service Provider switch, and the associated cost and complexity of configuring dynamic routes throughout a large heterogeneous network makes Layer 3 MPLS less attractive to deploy in the Metro.

Implementing MPLS at Layer 2 provides a less complex approach to VPNs than Layer 3 MPLS. Because Layer 2 MPLS VPNs create a separate virtual Layer 2 domain per subscriber, subscriber switches operate entirely at Layer 2 and Service Provider switches need not maintain any subscriber routing information. 


The Evolution to Multidimensional Ethernet

Multidimensional Ethernet is the next evolutionary stage of Ethernet required to reliably deliver real-time services in the Metro. By directly addressing the scalability and performance bottlenecks of existing VPN technologies, Multidimensional Ethernet enables Service Providers to overcome the last hurdles preventing Ethernet from completely taking over the Metro.

Multidimensional Ethernet enables next-generation Metro networks by extending the performance of Ethernet through a handful of critical technologies: Scalability with MAC-in-MAC, Service Density with Hierarchical Quality of Service (QoS), Ethernet Cross-Connect, and carrier-class Service Resiliency. With Multidimensional Ethernet, carriers and Service Providers will be able to increase capacity, improve quality of service, and reduce overall operating expenses. 

MAC-in-MAC Scaling: The first component of Multidimensional Ethernet is the proposed IEEE 802.1ah Provider Backbone Bridges standard, also known as MAC-in-MAC. The MAC-in-MAC name comes from how the standard encapsulates Ethernet frames with a Service Provider MAC header (see Figure 2). Through this frame format, it is possible to successfully separate Service Provider resources as the MAC-in-MAC encapsulation effectively tunnels subscriber traffic through carrier networks without any required interaction between the subscriber and provider.

For Metro applications, scaling is one of the most critical issues. MAC-in-MAC technology overcomes the inherent scalability limitations of VLANs and Q-in-Q networks that make them impractical for use in larger networks by enabling up to 4000 times as many service instances as supported by traditional VLAN and Q-in-Q networks. This provides an economic way to scale to millions of VPNs without increasing network complexity. Additionally, the MAC-in-MAC standard eliminates the need for core and backbone switches to learn hundreds of thousands of MAC addresses. Because the switches at the edge encapsulate the traffic with a service provider MAC address, the other switches in the core need only learn the MAC addresses of the core switches, as opposed to potentially hundreds of thousands of MAC addresses from attached devices. In effect, MAC-in-MAC increases scalability and reliability while reducing network complexity compared to other technologies.

Especially important, MAC-in-MAC implementations offer seamless interoperability with existing VLAN and Q-in-Q networks (see Figure 3). In this way, carriers can offer multiple VPN services to subscribers in a manner that most efficiently utilizes network resources, depending upon network topology, location, and subscriber density. Such hybrid architectures place less strain on the Metro, are less costly to provision, and require less support to maintain.

Hierarchical Quality of Service: The next critical element of Multidimensional Ethernet is Hierarchical Quality of Service. With the advent of Multidimensional Ethernet, new algorithms exist to handle the different requirements of business subscribers. Business subscribers typically negotiate contracts on an individual basis, and have much more dynamic requirements around quality of service for their varying applications. During the web-cast of a company training video for example, voice over IP traffic that normally has plenty of bandwidth could be interrupted if priorities are not mapped correctly. 

These new priority based algorithms allow the subscriber IT manager to set the priority of the various enterprise applications and the service provider can simply honor those settings within a fixed bandwidth profile across the VPN. The user no longer needs to decide the exact bandwidth requirements of each application. Within the bounds of the overall bandwidth profile, the service provider simply delivers the highest priority traffic first. When the highest priority applications are idle the lower priority applications are allowed to send their data (see Figure 4). This greatly simplifies the planning process for the customer, and also makes it easier for the network operator to market and sell business VPN services.

Ethernet Cross-Connect Flexibility: In the business service market, there is stiff competition to provide service and content streams to subscribers. For example, VoIP services can significantly reduce operational overhead compared to leased line implementations. Essential to the success of any network operator is the ability to provide access to best-of-class services and content, giving subscribers a full range of choices. Enabling such flexible provisioning in an efficient manner has, up to this point, been extremely difficult to implement. 

Rather than having to provide a separate pipe to service each market, with Ethernet Cross-Connect technology, multiple content providers can be accessed by different subscribers all through a single network connection at the business site. Ethernet Cross-Connects handle service connections at Layer 2, enabling flexible and simple to manage provisioning between multiple content provider networks. Additionally, this functionality significantly lowers deployment costs when compared to traditional router based implementations, yielding a tremendous return on investment.

For a business that uses different VLANs per site, Ethernet Cross-Connect technology allows the metro service provider to translate the customer tags between the sites, so the business can leave its internal infrastructure unchanged through mergers and acquisitions and other activities that require convergence of multiple networks. Figure 5 shows an example of this capability.

Carrier Class Resiliency: The cornerstones of a profitable Metro network are Quality of Service and performance. Without reliability, however, even the highest performance network will be unable to service customers with the guaranteed levels of service they've come to expect from carrier-class circuit-switched networks. In order to achieve "five nines" reliability, Multidimensional Ethernet must do more than offer a wider range of services with better QoS.

For these reasons, the final component of Multidimensional Ethernet is Service Resiliency. Service Resiliency provides failure avoidance and recovery mechanisms to the applications and services supported by the network. Rather than just guaranteeing that a broken link will be quickly restored, Service Resiliency guarantees that every service carried over a link will be maintained and restored as well.

Multidimensional Ethernet ensures Service Resiliency through Ethernet Automatic Protection Switching (EAPS):. The EAPS protocol has been published as IEEE RFC 3619, and is a widely deployed protection mechanism for native Ethernet interfaces which is compliant with the MEF 2 (Metro Ethernet Forum) Service Protection Framework. EAPS utilizes a standard Ethernet MAC and a ring topology to provide service-aware protection with carrier-class failover response within 50 ms (See Figure 6). EAPS also promotes route diversity to maximize spatial reuse by enabling Service Providers to select primary and backup designations on a per-VLAN basis. The redundancy built in to the hardware, software and network layers of Multidimensional Ethernet offers true Service Resiliency to the business subscriber.

Next-generation Metro Ethernet networks require a high level of performance, flexibility, and QoS granularity while reducing system complexity and cost. Multidimensional Ethernet, with its four cornerstones of MAC-in-MAC scaling, Hierarchical QoS, Ethernet Cross-Connect flexibility, and Service Resiliency, is the key technology enabling carriers to deliver a successful -- and profitable -- rollout of converged business networks.