Let’s start by recapping Ethernet’s attractions. In a nutshell, it lowers the cost of ownership for the customer and the carrier, paving the way toward broadband packet services that can be offered cost-effectively to the broadest customer set, over a variety of network infrastructures.
I previously argued that Ethernet combined with Multiprotocol Label Switching (MPLS) can eventually supplant frame relay as the dominant enterprise WAN packet service.
This will take time, as it requires major carriers to deploy MPLS core networks that will represent a new distributed switching fabric, augmenting or replacing frame and Asynchronous Transfer Mode (ATM). Along the way, carriers also need to think about how to get users onto those networks. Today, frame relay is typically provided via T1 leased-line access to a local provider’s frame relay access point.
As the transition to Ethernet-based networking occurs, it will be important for carriers to provide high-speed Ethernet-based access to their service POPs (points of presence). Today, that can only be accomplished via optical connections, typically Gigabit Ethernet transport between a building basement or enterprise data center and a POP.
That’s the undoing of many an “EtherLEC” (Ethernet local exchange carrier), as getting that fiber optic lateral to a building is costly or just plain impossible, thanks to right-of-way access limitations and/or intransigent building owners.
That’s why these days, when you talk about Ethernet services, you end up talking about copper. If fiber is the problem, then just get around it. Fixed wireless access has been offered as an alternative, but this requires wholly new systems to be deployed, and so far everyone’s experience with broadband wireless in the U.S. has been poor.
The focus among all the big carriers I’ve talked with has been wringing as much revenue as possible out of buildings already on their networks, not building out those networks to new buildings. Expanding the capacity of their copper plant is key to this mission.
Today, approximately 75 percent of all businesses and homes are within one mile of a central office or DLC (digital loop carrier), so expanding the carrying capacity of that copper is the easiest way to introduce any new broadband service. Homes are doing well enough with ADSL, so the focus of EFM (Ethernet in the First Mile) is primarily on first-mile access to buildings over copper, or in-building distribution of Ethernet over twisted-wire-pair copper.
Much of the work in getting Ethernet happily onto copper is happening within the IEEE's 802.3ah Ethernet in the First Mile Task Force. Within this group, silicon and systems vendors are working on defining the specs that will shape how Ethernet makes its way into the local loop and up the riser.
Simply put, the EFM Task Force wants to see standards develop for solutions that deliver duplex 10-Mbit/s Ethernet over copper loops up to 750 meters. There are many other rates and reaches proposed, but this captures the essence of what Ethernet-over-copper (EFM/Cu) is after.
The EFM task force is tackling lots of other issues, but copper is arguably the most interesting, and contains within it the most contentious debates and dazzling promise of riches (millions of access lines!). Here’s a sampling of what’s taking place with EFM/Cu lately:
Business vs. Residential
There are proposals for different rate and reach schemes for business and residential markets. As the concept of EFM evolves, so does the pressure from carriers for different standards for different applications or customer groups. In a presentation from Sprint Corp. and SBC Communications Inc. at a recent meeting of the EFM Task Force, the two carriers argued for two distinct rate and reach objectives for EFM over copper.
For the residential market, they argued that there should be a specification for EFM/Cu that is asymmetric; optimized for delivery of voice, video, and data to homes; and compliant with American National Standards Institute (ANSI)'s T1E1.4 Band Plan 998.
For business customers, the EFM/Cu PHY should be symmetric, and optimized for data services only. Proposals today include the VDSL Band Plan 997 and an enhanced version of G.SHDSL.
A religious war continues to rage over whether the line encoding scheme for Ethernet over VDSL (very high bit-rate DSL) should be QAM (quadrature amplitude modulation) or DMT (discrete multi-tone).
Hard to believe this war is religious, but it is. On the one side you have the chip vendors from the cable TV and modem world pushing QAM as the simple and easy way to get lots of bandwidth onto copper twisted-wire pairs. Vendors like Broadcom Corp., Infineon Technologies AG, Metalink Ltd., Paradyne Networks Inc., and Tioga Technologies Ltd. (plus a lot of Asian chip vendors) believe that Ethernet over QAM-VDSL will win by its advantages of low cost, simplicity of design, low latency, plug-and-play operation via fast blind synchronization, and low power consumption.
The VDSL-DMT supporters, including Alcatel SA, GlobeSpan, IBM Corp., Ikanos Communications Inc., and Zarlink Semiconductor Inc., are arguing that since they carried the day in the ADSL standardization war years ago, they should be given the crown in VDSL as well. According to recent presentations to the IEEE, VDSL-DMT supporters say digital complexity in DMT scales easily with smaller semiconductor process geometries, while QAM has more analog complexity due to filtering, which does not scale cost-effectively. Additionally, no QAM solutions on the market today are fully standards compliant.
Symmetric or Not
Some are beginning to argue that VDSL is ideally suited for asymmetric service delivery, such as bundled voice, video, and data services to homes, whereas business services are best supported by symmetric DSL solutions. Additionally, VDSL is severely limited by loop length, making it less suitable for networks with little fiber outside the feeder.
An “enhanced” SHDSL (symmetric high bit-rate DSL) for EFM/Cu is currently being explored by the International Telecommunication Union, Standardization Sector (ITU-T) and ANSI T1E1.4. This would be better suited for longer copper loops and business-oriented data services that require symmetric bandwidth. It follows from earlier symmetric DSL technologies including HDSL, HDSL2, and SDSL.
The performance currently under consideration is 5 Mbit/s at 3,600 feet (or 1.2 km per pair) and 3.33 Mbit/s at 5,000 feet (or 1.6 km per pair). This enhanced SHDSL is achieved by increasing the number of bits per symbol, and utilizing bonding techniques to combine the transmission speeds of multiple copper pairs. It’s also worth noting that there is a growing feeling that if the religious war between QAM and DMT can’t be resolved, enhanced SHDSL may just win, because carriers aren’t invested in line coding wars at all – they just want results.
Looking at the few remaining hot spots in networking – converging the data layer with MPLS, converging the transport layer with NG-Sonet (GFP/X.86), Ethernet-based networking and service infrastructure – EFM/Cu fits in nicely, extending the ubiquity of Ethernet to residential markets and to a broader addressable market for business services. The chip vendors are going to make the most noise about this, because they are the first enablers of this solution, so they've a lot to gain and a lot to lose.
An interesting debate going on in the venture community is where the highest value of Ethernet-over-copper is located: in the chip or in the system. I’m inclined to think the chicken always sells for more than the egg, so my bet remains with the system vendors (though chickens can cost a lot to feed, can’t they?). The question many VCs will be asking this year is whether to bet on the next Broadcom or the next Alcatel.
What is likely to be another bruise to our national pride is the speed at which this technology is advancing in Asia and Europe, where the deployment of next-gen infrastructure is more closely tied to political will than to the will of Wall Street. This time next year, we’ll likely be looking to Korea, Japan, and China for wisdom from their experiences deploying this gear. The independent LECs in the U.S. are starting to dabble in VDSL, but it’s nothing like the scale of deployments found in Asian countries today.
SBC remains serious about expanding its access offerings to include EFM, and even IXCs see the value in expanding the range of revenue-bearing services they can offer to buildings currently on their networks. What we’ll look at next is just what happens in the core of the network once thousands of users are accessing that network via Ethernet. Even the most graceful migration to a new access method has widespread implications for the core, and Ethernet will clearly be driving those over the next decade.
— Scott Clavenna, Director of Research, Light Reading