Conventional wisdom and marketing buzz both say that enterprises should replace existing ATM and frame relay services with Ethernet. This could very well be the case, providing network planners can identify a carrier that meets a number of key criteria. But a nagging question lingers: why would a company be prepared to retire a functional service‾
Two of the strongest motivations for such a network transformation are to get high-speed performance, with up to 10-Gbps speeds not unrealistic, and to lower operational costs. Ethernet services can provide better price-per-bit ratios compared with legacy solutions, as well as reduced costs in sourcing leased lines and deploying routers with wide-area network (WAN) interfaces.
On top of these benefits, the Ethernet switches that are ubiquitous in enterprise environments provide the convenience of plug-and-play. Additionally, where speed is a major buying criterion, Ethernet is excellent for dedicated Internet access, data center connections and LAN-to-LAN services where quality of service (QoS) and low latencies are especially meaningful. Finally, in terms of convergence, the Ethernet pipe is capable of carrying voice, data and video, delivering a wide range of opportunities for the use of innovative applications.
So with all this sunshine, where are the clouds lurking‾ One of Ethernet's main inhibitors is its dependence upon fiber links. The introduction of low-speed, 12-Mbps mid-band Ethernet-over-copper-pair is gaining momentum thus offering budget-conscious prospects, such as schools and municipalities, a realistic approach.
In order to fully replace frame relay and ATM as the new enterprise backbone, however, carrier Ethernet needs to:
- Evolve into a service that supports a switched any-to-any architecture
- Become independent of physical-layer transport
- Become distance insensitive
- Provide scalability to national and international proportions.
Furthermore, efficient recovery mechanisms are essential, along with traffic management and QoS that are on par with legacy and next-generation VPN services. Some carriers can provide these attributes, but not all, so prospective buyers must ask these questions before embarking on any network evolution project.
Most of the desired attributes can be met, for example, by MPLS-based IP VPN services as specified by IETF RFC 4364 (formerly RFC 2547bis), but some of the largest frame relay and ATM enterprise networks are in fact migrating to Ethernet, not IP. Why‾ Enterprise IT staff are not willing to turn their routing tables over to the carrier. Instead, this type of company will control their own IP routing over an MPLS-based Ethernet VPN. This type of Layer-2 VPN is known as a virtual private LAN Service (VPLS).
In a VPLS, MPLS pseudowires provide logical separation and control of customer traffic similar to frame or ATM virtual circuits, yet the network can support the meshed topologies that confound legacy Layer-2 solutions. MPLS also supports the traffic engineering capability necessary for distinct QoS. And, finally, MPLS has a fast reroute recovery mechanism that is far superior to Ethernet's Spanning Tree protocol.
The only thing holding up mass migration is a lack of VPLS from top-tier carriers.