Ethernet in the Metro and the Advisability of Placing Ethernet Switches in the Central Office

Most of the wireless broadband equipment adhering to a metro Ethernet model conforms to
the 802.11 standard rather than to 802.16. The 802.11 equipment is Ethernet based, and 802.16
is IP based; that’s the distinction.
Ethernet is a layer-2 protocol rather than layer 3; in other words, it employs circuit switching
to direct traffic over the network instead of packet routing despite that Ethernet itself is a
pure packet protocol. Switched Ethernets divide the overall network into subnets, generally
based on the geographical location of the terminals, and the switch sends traffic to the appropriate
subnet. To see how Ethernet switches fit into a public network, refer to Figure 6-2. I am mentioning Ethernet switches here in the context of the central office because a considerable
body of opinion says that Ethernet rather than IP is the proper protocol for the MAN.
In fact, no fewer than two industry associations, the Metro Ethernet Forum and the Gigabit
Ethernet Alliance, have arisen to promote such views.
Within the telecommunications world, a high degree of partisan fervor often accompanies
advocacy of one protocol or the other, and indeed the stakes are high and are nothing less than
market success or failure for the party embracing one standard or other. A dispassionate consideration
of all relevant transport protocols is far preferable to rabid partisanship, and the
criteria on which a transport should be judged are the cost effectiveness and flexibility of the
equipment in supporting a given service model as well as the likelihood that the standard itself
will gain wider currency in the future.
That said, you should consider the past and future of metro Ethernet: High-speed metro
Ethernet business class services were introduced by several competitive data services in 2000,
including offerings from Yipes, Cogent, Terabeam, and CAVU-eXpedient, among others, the
last two of which used wireless infrastructure. The initial acceptance for such services was
encouraging, but two, CAVU-eXpedient and Yipes, declared bankruptcy, and neither Cogent
nor Terabeam has succeeded in posing major challenges to the frame relay and T1 incumbents,
though they have managed to survive. Recently large long-distance companies as well as
local incumbent telephone carriers have also begun to introduce both IP and metro Ethernet
services, though as yet such services have very small subscriber bases.
The independent metro Ethernet service providers all followed a similar model, offering
simple high-speed access at high throughput rates and no value-added services whatsoever.
Basic Ethernet is well suited to such a model, but it is much less suitable to a more contemporary
model where the network is used to deliver converged services. An Ethernet substandard
for supporting virtual private networks (VPNs), 802.1q, does exist, as well as another, 802.3x,
which imposes flow control over the network. But far less standards work has been done for the
purpose of enhancing network performance with delay-sensitive traffic than in the case of IP. The general Ethernet solution for any problem is to throw more bandwidth at it rather than to
invoke sophisticated network management mechanisms, which explains the steady and rapid
increase in speed from 10 megabits per second (Mbps), to 100Mbps, to 1 gigabits per second
(Gbps), and finally to 10Gbps. Incidentally, a 100Gbps standard is in preparation.
Some companies, most notably Extreme Networks, have developed extensive proprietary
QoS protocols for Ethernet that greatly augment the capabilities of the Ethernet switch, but
one is still left with the question, why not use a standard-based approach such as IP/MPLS?
In a book of this sort, any exhaustive review of the arguments pro and con for Ethernet in
the metro would be inappropriate. The fundamental argument is that Ethernet-based components
will be cheaper because of mass production of products intended for the enterprise and
residential markets. Certainly, some of the same chipsets can be used in carrier products as in
less-demanding applications, but a degree of redundancy and intelligence must be provided in
a product intended for a public network, and such requirements will serve to elevate the price.
The relatively few metro Ethernet products to appear thus far have not been inexpensive.
A second argument has it that enterprise managers, the individuals who will actually be
using broadband services, are happiest when interfacing to an Ethernet that is simply a larger
version of what they themselves manage. There is some truth to this argument. But given the
lesser capabilities of Ethernet in the area of service creation vis-à-vis IP/MPLS or ATM, is this
singular virtue sufficiently compelling?
My view is that the momentum behind metro Ethernet has significantly lessened. Furthermore,
only two companies, Atrica and Extreme Networks, currently manufacture pure carriergrade
metro Ethernet switches (numerous companies make multiservice-switching platforms
that can do Ethernet switching as well as routing). The whole notion behind the original metro
Ethernet movement of simply offering high-throughput speed to a subscriber and leaving the
subscribers to establish their own services strikes me as largely discredited today. Still, one
never knows. Perhaps some service provider will make a success with this model yet.
Also, an Ethernet switch can be combined with an MPLS switch with or without including
full IP routing capability. Products from Atrica and Extreme Networks are essentially MPLSenabled
Ethernet switches, and Laurel Networks, TiMetra (now a division of Alcatel), and
Vivace (now a division of Tellabs) make boxes that can function as Ethernet switches, IP routers,
and MPLS switches. None of these products is specifically designed to operate within a
wireless broadband network, though, and interface problems may arise. Incidentally, all these
companies are startups that tout at least partially proprietary approaches to service delivery,
and all their products are expensive and intended for inclusion in larger networks.
Finally, the inclusion of an Ethernet-switching capability at the central office may render
a wireless service offering more attractive to certain customers. Here again, it is good to assess
the needs and profiles of potential subscribers before proceeding with the construction of
the network.