The original Ethernet standard, attributed to Robert Metcalfe and David Boggs, was developed in the 1970s as a LAN technology that allowed multiple users to share a common high-speed cabling backbone. Ethernet has subsequently been reinvented many times over the years, making the transition to new media such as fibre and unshielded twisted copper, as well as achieving incremental increases in its operating speed from the original 10Mbit/s through to multi-gigabit-per-second performance.
The increased data rate achieved over longer distances can be mainly attributed to the fact that Ethernet can now adopt a physical point-to-point topology, where a given link offers dedicated bandwidth to the sending and receiving interfaces. This is in marked contrast to the earlier implementations of Ethernet, where a shared medium meant collision detection and re-transmission mechanisms had to be implemented that imposed strict distance limitations.
In recent years, Ethernet has moved into the wide-area domain, which seems a rather bizarre development at first glance, given the origins of the standard. The main driver behind this trend is the ubiquity of the Ethernet interface – almost every networked device today, from a low-specification PC to a carrier-class router, comes equipped with an Ethernet port. As a result, Ethernet can justifiably be viewed as the de facto choice for the lower network layers.
Ethernet offers transparency to any Layer 3 protocol, making it unnecessary for changes to be made to existing network protocols that may be in use. Also, because Ethernet is ‘connectionless’, it can provide logical multipoint-to-multipoint connectivity without any special configuration, making it especially suitable for media broadcasting applications. Finally, the security of individual VPNs can be assured in a carrier network by attaching VLAN tags to each Ethernet frame.
Ethernet in the wide-area environment is of interest not only to enterprise users, but can also be used for applications such as delivering triple-play services to domestic users and carrier applications such as backhaul connectivity for wireless networks.
Ethernet used for metro and wide-area applications is generically described as Carrier Ethernet. There are, however, a number of variants of Carrier Ethernet that have evolved during the past ten years or so, each bringing progressive performance and cost advantages. Since their inception, Ethernet LANs could be interconnected between geographically dispersed locations, but this was achieved by disassembling the Ethernet frames and transmitting the data in a wide-area format prior to reassembly at the far end. This process was inefficient because the payload management overhead consumed a significant amount of the available bandwidth, and the process was technically complex and expensive. Legacy Ethernet standards do, however, lack the sophistication of the traditional wide-area protocols – especially in the areas of detecting and recovering from link failures and, of particular importance to carriers, operational management.
Today, Carrier Ethernet can be broadly divided into two distinct areas: first, Ethernet as a carrier service, whereby the carrier offers an Ethernet interface to its customers and subsequently pass the service over another transport layer as described above; and second, Carrier Ethernet transport (CET), whereby Ethernet is used as an internal transport mechanism in its own right across the carrier’s network, which is arguably the ‘true’ Carrier Ethernet.
Ethernet as a carrier service can use a number of Layer 1/2 protocols in the core, typically SDH, DWDM and optical transport. In recent years, the emphasis has been mainly on using IP-based carrier networks running MPLS, which can be used to transport Ethernet. This is referred to as Ethernet over MPLS (EoMPLS), where MPLS provides the necessary QoS, and is attractive to carriers because they can use their existing IP-based infrastructure to act as the bearer for Ethernet services and the maturity of the technology means that it offers a stable platform. The IETF has developed two important standards around EoMPLS: Virtual Private Wire Service (VPWS) and Virtual Private LAN Service (VPLS), which provide point-to-point connectivity and any-to-any connectivity, respectively.
CET is the area in which some of the most interesting developments are taking place. The standards development work has been mainly undertaken by the IEEE and ITU – the latter’s involvement reflecting the bias towards a carrier-based standard, although such distinctions are becoming increasingly blurred.
There have been a number of threads in the standards development activity, most of which has focused on solving the inherent weaknesses exposed in Ethernet when it is transferred to the carrier environment.
- Resilience: LANs have traditionally ‘self healed’ by using the Spanning Tree Protocol to reroute traffic in the event of a primary path failure, but this mechanism is not fast enough to provide the rapid recovery demanded in carrier networks – that is, about 150ms compared with the sub-50ms associated with SDH. Resilient Packet Ring (RPR) subsequently provided sub-50ms recovery in the metro ring environment and the later development of Provider Backbone Bridge Traffic Engineering (PBB-TE), which facilitates rapid recovery by using a deterministic reroute controlled from a configuration database, has brought further sophistication for carrier-scale networks.
PBB-TE is also being further developed to provide bandwidth and QoS guarantees.
- Separation of carrier and customer networks: CET can only be delivered effectively if the customer’s Layer 2 or 3 network is logically separated from that of the carrier. This is achieved by providing separate VLAN tags and, potentially, separating the MAC addressing as well.
- Manageability: CET needs a carrier-grade management capability to be acceptable as a standalone protocol for use in core networks. This has been addressed with Ethernet standards such as OAM 802.1ab and ag and further developments are likely in this area.
If Ethernet can reach a sufficiently mature status to support core carrier networks, it will replace earlier and more-expensive standards such as SDH. This is predicted to deliver capex savings of about 50% and also reduce opex because of the simplified nature of the network architecture. Its deterministic connectivity across the core network solves the latency issues that have tarnished the reputation of early generations of Ethernet as a service offering.
In conclusion, CET promises to become the final frontier for delivering Layer 2 connectivity in the local and wide-area environments for multiple applications, ranging from enterprise data to domestic triple-play services. It will deliver both cost and performance benefits in comparison with legacy solutions, and reduce the complexity of WANs. Most carriers have yet to migrate to CET, but several new entrants have already launched fledgling CET networks.