One of my favorite Mark Twain quotes is a riddle about semantics: If you call a dog’s tail a leg, how many legs does the dog have? The answer: Four, because calling a dog’s tail a leg doesn’t make it one.

Semantics can be tricky when it comes to technology, too. Take the concept of “carrier Ethernet.” We think we know what it means: Ethernet services provided by carriers. But it’s not that obvious. There are three distinct offerings, all with specific characteristics, under the broad umbrella of carrier Ethernet.

Worse, these days all three offerings are typically provided over the carrier’s MPLS backbone, so technically, all three qualify as “MPLS services.” (Remember, MPLS is a technology, not a service. It can be used to offer a range of services, from Layer 1 through Layer 3. Yet confusingly, both users and carriers refer to Layer 3 MPLS, defined in IETF RFCs 2547 and 4364, as “MPLS.”)

The three flavors of carrier Ethernet are:

— Ethernet access to MPLS services. Here, the carrier provides a direct Ethernet interface to “classic” Layer 3 MPLS services. The user’s customer edge (CE) router connects with the carrier’s provider edge (PE) router across an Ethernet interface, and the carrier routes the user’s traffic across the cloud. The upside: higher-speed access to traditional MPLS services, with all the QoS support that MPLS provides. The downside: the carrier is involved in IP-layer routing.

— Point-to-point Ethernet services. Here, the carrier provides a direct Ethernet interface to the user, but only at Layer 2. The user’s CE router connects to a carrier switch (typically MPLS) that transports traffic across the cloud, but looks and feels like a WAN Ethernet link. That is, the IP device that the user’s CE router is connected to is another CE router across the cloud. The upside: a clean, simple link across the WAN, ideal for, say, high-bandwidth links between data centers. The downside: as the name implies, the link is point-to-point, meaning that the service doesn’t scale to a full-blown WAN.

— Multipoint Ethernet services, or virtual private LAN services. Here, the carrier provides a direct Ethernet interface to the user, again at Layer 2. The user’s CE routers interconnect across what looks, smells and feels like a “LAN across the WAN” — with all IP routing controlled and managed by the user. The upside: it’s simple to install and configure, compared with MPLS. The downside: QoS isn’t native (though carriers have different ways of prioritizing customer traffic at Layer 2). Additionally, there may be some challenges with scalability. I haven’t seen any multipoint Ethernet networks larger than roughly 200 sites — MPLS networks, in contrast, are often several thousand sites.

Users are increasingly interested in adopting all flavors of carrier Ethernet — 63 per cent of the folks I work with have some flavor of Ethernet services, and 85 per cent say they plan to expand their use of such services, primarily due to the ease of configuration and lower cost of bandwidth. The key to doing so successfully? Know which one you want.