The core of the Internet is awash with terabits of bandwidth, and metro rings are coming up everywhere. The networks that provide subscribers access to this huge bandwidth are woefully inadequate, both in terms of availability and bandwidth. There is a huge set of services waiting to be offered to the customer, services like streaming video, digital TV, high-speed Internet access, etc, but the access network is holding up this business. An example of the bandwidth required by these services vis-à-vis the available access bandwidth, multiple digital TV requires access rate of up to 24 Mbps, whereas most subscribers are stuck at sub-megabit speeds.

Why Ethernet?
Existing access solutions are inadequate in one respect or another. DSL, cable modems and ISDN offer inadequate data rates, satellite is uni-directional, broadband wireless solutions are not yet mature enough. In short, all existing solutions have some or the other problems associated with them.

Given that lack of subscriber bandwidth is holding up the growth of the Internet, leading vendors are working on providing an easy-to-implement metro access solution, one that will bridge the bandwidth gap. Surprising as it may seem, the protocol that made a humble beginning as a ‘Me Too’ LAN protocol is poised to become the layer 2 protocol of choice at the metro access. While the reasons for Ethernet rising above all others are many, there is little doubt that Ethernet-to-the-X vendors are taking it seriously. First, Ethernet is universally available—there are over 300 million Ethernet cards installed today. Second, with such huge volumes, economies-of-scale have driven down the costs. Third, the technology is stable and widely known, and a highly skilled and well-trained talent pool is available to install and administer Ethernet networks. Fourth, Ethernet lends itself to various types of media, with little or no change to the MAC layer, thereby allowing existing applications to talk to each other without re-work.

Before we get into the details of the various Ethernet solutions, we need to study the peculiarities of the metro access network, the topologies and the subscriber scenario’s.

ETHERNET ON MOVE
	The task force addressing these issues is the IEEE EFM working group, which is moving towards standardization of the 802.3ah specifications
	The issue is being driven by big names like Cisco and Intel, and so one is likely to see some action soon

The figure above explains the basic deployment scenarios and the associated media/physical layers’ characteristics of the entire Ethernet metro access solution. In order to emphasize the fact that the Internet places top priority to the customer, the subscriber loop is called the ‘first mile’ (as opposed to the ‘last mile’ of the core-centric telecom world), and we shall hereafter refer to it as Ethernet in the first mile (EFM).

No single EFM solution can be deployed in all subscriber scenarios. This is chiefly because the requirements of access subscribers vary, and because there is a need to use the existing media infrastructure. To accommodate these constraints, let us first study the types of access subscribers, as well as their particular needs.

The subscriber could be a business, a residence, a large campus, hotel or hospital. The residence itself could be a series of single-dwelling units, or a multi-tenant apartment block. Each of these subscriber scenarios impose their own constraints, and there is no single solution that can fit in all these scenarios. The issue is further complicated by the fact that there could be diverse media, and fiber may not be available in the basement or at the curb. Investment in the existing copper cabling could be so great that the infrastructure provider may not be able to discard his wiring. EFM is therefore not a solution, but a solution-set that addresses all these complications.

Broadly speaking, EFM addresses two types of media, two different network topologies, and three different physical layer specifications. The task force addressing these issues is the IEEE EFM working group, which is moving towards standardization of the 802.3ah specifications. These groups are being driven by big names like Cisco and Intel, so one is likely to see some action soon.

Adapting Ethernet in the metro access centers on defining physical and MAC layer specifications for each of the network topologies given in the figure above, while leaving the LLC untouched. This means that, given a little encouragement, all existing Ethernet-based devices and applications running these devices would work seamlessly with this solution. Since the most widely available media in the metro are copper and fiber, only these have been targeted. Also, we observe that point-to-point and point-to-multipoint network topologies are being worked upon, and the choice of one or the other is dictated by bandwidth requirements, subscriber density/geographic distribution, cost, and media availability. Let us study each of these solutions, and explore the scenarios where they are likely to be deployed.

Point-to-Point over Copper
This technology is the way to go where required data rates are less than 15 Mbps, a large installed copper base exists, and dstance to the metro rings is less than a mile.

This solution is likely to be deployed in MxU’s and campuses. MxU is a generic term for large apartment complexes (multi-dwelling units), commercial complexes (multi-tenant units), hotels (multi-hospitality units), etc. The services that would be offered would typically be convergent services like voice and video and high-speed Internet access. This technology would typically be chosen where initial investment is constrained; quick installation is a pre-requisite and reasonable speed is desired.

The EFM sub-group at the IEEE that is looking into defining a suitable physical layer specification has not yet converged on the physical layer standards that need to be adopted for Ethernet over copper. The current trend seems to be veering towards adopting a compromise between the ANSI VDSL PHY and the SHDSL PHY. In the meanwhile, vendors are pressing ahead with their development.

Point-to-Point over Optical Fiber
Where fiber is available or future proofing is required, point-to-point Ethernet over fiber would be a good choice.

Ethernet over fiber offers a new business model, and a large set of revenue-generating services, tremendous scalability in terms of bandwidth and subscriber density, and an extremely flexible service provisioning model and longevity of infrastructure. While all of the above are properties of optical fiber, Ethernet as a layer 2 protocol adds a new dimension ie low-cost 1000BaseX cards, which have been in production for years now.

Service providers who are looking towards installing a new network infrastructure will find this an interesting choice, for aforementioned reasons. The lifetime costs of the network would be reduced greatly, while ensuring a high degree of protection against obsolescence, though initial investments could be high.

Point-to-Multipoint over Fiber
While P2P solutions over copper offer a cost-effective, low-investment, high-speed access to the metro network, the level of services that this can carry is limited. For instance, a multi-channel digital TV requires 24 Mbps per subscriber, while EoVDSL delivers up to 15 Mbps. The alternate solution, point-to-point over fiber, solves the problem of bandwidth, but involves high initial investments, and delivers more bandwidth than most subscribers require (up to 1 Gbps). Ethernet passive optical network (EPON) solves this problem. It provides reasonable amount of bandwidth at low costs. EPON can deliver bandwidth of up to 30 Mbps, and could provide higher bandwidth for short periods of time (a subscriber could, for instance, ask for and obtain a burst of 75 Mbps bandwidth for downloading large amounts of data, and revert back to 30 Mbps when done).

In addition to delivering adequate bandwidth at low costs, EPON offers several technical and business advantages to the infrastructure provider. Instead of drawing a fiber from the service provider PoP to each customer premises, the fiber is pulled till a convenient point close to a cluster of subscribers, and an optical splitter is used to transport data to subscribers from that point onwards. This not only saves the length of fiber to be installed, but also reduces the density of the fiber to be managed at the PoP. This results in lesser number of transceivers in the central office, lower power requirements, lower space requirements, and lower operations and management costs. Because field devices are passive in nature (of the order of one per 32 customers ), they do not need power and are free of electronics. This translates into significantly low outdoor plant costs and maintenance costs.

Implementation Issues
However, it’s not roses all the way. Since an Ethernet network was to be administered by a single entity confined to a single LAN, the protocol did not define the elaborate operation and maintenance mechanisms, which are vital to running large networks. Nor are there any robust self-recovery mechanisms. To address these issues, the EFM working group has formed a sub-group for the purpose of defining an operation, administration and management (OAM) scheme to enable metro-wide Ethernet networks to be managed. The importance of this group cannot be emphasized enough, suffice it to say that Ethernet could be well on it’s way to becoming a true carrier-class protocol if large networks can be managed and administered remotely.

There seemed to be two schools of thought about how to embed OAM into Ethernet. One school was for defining a special MAC layer OAM data unit, the other was for using the PHY preamble (largely redundant) to carry OAM data. The group seems to be veering towards the preamble solution, being pushed by the big guns in the industry.

In all, Ethernet in the first mile offers a broad range of solutions for all needs, reduces costs, increases bandwidth, and is poised to be the engine of growth for the Internet. In these lean times, this could mean the difference between life and death. All the three solutions will be deployed in some measure or the other though one may suspect if EPON has enough merit to warrant special attention.

The day when a humble LAN protocol runs the Internet is not far.

Arif Shouqi, Cisco Systems Global R&D center