Ethernet: Poised to go the distance

Communication service providers constantly look to bring more services to subscribers at the least possible cost. The recently approved Ethernet in the First Mile (EFM) standard offers providers a way to make the "triple play" dream of voice, video, and data available to everyone at a reasonable cost. Fortunately, testing EFM isn't much different from testing other wireline communications technologies.EFM defines new copper and fiber interfaces for Ethernet along with new operations, administration, and maintenance methods for managing Ethernet subscriber networks. The Ethernet in the First Mile Alliance (www.efmalliance.org) claims that EFM will bring Ethernet from every desktop and local network to the weakest link on the communications chain—the first mile. Also called the "last mile" or the "local loop," the first mile resides between telecom and cable service providers and consumers (homes and businesses). The technology gained formal recognition when the IEEE approved it as IEEE 802.3ah in June 2004 (Ref. 1).

Ten years ago, the Asynchronous Transfer Mode (ATM) protocol burst on the scene with promises to bring digital voice, data, and video to everyone. A layer-2 (data-link layer) protocol, ATM grew to dominate the electrical and optical links that connect home computers and business networks to multiplexers that provide access to public networks. Since its inception, ATM has had the ability to guarantee quality of service (QoS) through a sublayer called Operations, Administration, and Maintenance (OAM).

Ethernet, a simpler protocol designed for just data communications, became king of the local network. But Ethernet lacked the OAM capabilities needed to bring sufficient QoS to carry voice and video in the local loop. That's about to change. With the recent approval of 802.3ah, EFM is now poised to replace ATM over local loops. Service providers should see lower capital expenditures with EFM because their networks won't need equipment to convert between the ATM layer-2 protocol and the Ethernet two-layer protocol.

Figure 1. Ethernet in the first mile could replace ATM between subscriber premises and a service provider.

IEEE 802.3ah specifies that EFM will use a DSL-based copper link or either of two fiber-optic links in the local loop. Table 1 highlights the distances and throughput rates of the three physical layers.

The copper link can use either discrete multitone modulation (typically used in ADSL links) or quadrature amplitude modulation (typically used in cable-modem links). EFM-based modems won't need the protocol stacks that today's modems need to convert ATM cells to Ethernet frames because ATM won't exist in an EFM loop.

IEEE 802.3ah specifies two types of fiber-optic links: point-to-point (P2P) and point-to-multipoint (P2MP). Large businesses or campuses might use P2P connections, while small businesses and homes with fiber access might connect through P2MP links.

a)
b)
	Figure 2. Ethernet passive optical networks use splitters to connect a central office to subscribers through (a) star and (b) tree topologies.

EFM differs from other forms of Ethernet because it includes an OAM sublayer within layer 2. OAM adds monitoring, testing, and reporting abilities that can bring a link's QoS to levels acceptable for voice, video, and data. The EFM Alliance has published a tutorial that explains how the OAM sublayers work (Ref. 2).

Figure 3. EFM adds an operations, administration, and maintenance (OAM) sub-layer to layer 2 of the protocol stack.

The link monitoring feature lets two network elements check the health of the link between them. Timers keep messages running between the two. A network element must receive a link-monitoring frame every 5 s to keep the link alive. Link monitoring checks for errored symbols/s, errored frames/s, errored frame periods, and errored frame seconds and can report those results to a network-management system.

A remote-failure indication prevents a total communication breakdown. In traditional Ethernet, if one communication direction fails, the link terminates in both directions. Not so with EFM. IEEE 802.3ah defines a protocol where if one direction fails, the other may continue to transfer failure messages to adjacent network elements. Thus, if one element's transmitter fails, the healthy element in the link can tell the failed one to stop transmitting and can notify a network-management system of the failure.

When configured for a loopback test, a network element such as a subscriber's modem returns frames sent to it from another network element such as a DSLAM, as long as both elements share a physical link. Service providers can use this feature to test new subscriber lines before placing them into service. "Remote loopback testing is the most important OAM feature in EFM," says Isabelle Moremcy, director of operations at Iometrix, a telecommunications test company and test-suite developer. "It lets you send traffic to a remote link and verify the continuity of the link."

As with any new communications technology, EFM devices require performance, conformance, and interoperability tests. At the physical layer, many of the tests are similar to those used in other technologies. For copper links, you'll have to measure signal-to-noise performance, bit jitter, receiver sensitivity, and crosstalk just as you do for DSL equipment (Ref. 3).

For P2P optical links, you'll need to conduct many of the tests that you already perform on Gigabit Ethernet links. Tests such as bit-error-rate, bit jitter, eye masks, stressed eye, optical power, and receiver sensitivity remain essential.

Yet, some aspects of EFM require tests that differ from those used for other Ethernet applications. Wael William Diab, technical leader at Cisco Systems and chief editor of the IEEE 802.3ah standard, notes that the IEEE 802.3ah specifies optical links at 10 km as opposed to 5 km for other Ethernet networks. Diab also notes that testing EPON links is more tedious than testing P2P links. Because EPON links can take on a myriad of topologies, they require more tests for optical power than P2P links do. The test topologies you need to use in test networks will depend on the applications you expect your products to fill.

For example, suppose an EPON link starts from the central office, splits off to one subscriber, then continues through a series of splitters off the main fiber to other subscribers (Fig. 2b). When you build a test network, the first splitter shouldn't divert 50% of the transmitted power to one subscriber. You'll need asymmetric splitters in your test network so all subscribers' equipment receives sufficient light power.

An EPON link also carries data from subscribers to the service provider, but not all of the light sent from one subscriber's equipment will reach the provider's equipment. Some light will refract onto other subscriber links. You must test your devices for immunity to frames carried on refracted light, which doesn't occur in P2P links. Be sure to test your devices with different splitters and numerous network elements connected to your splitters. Adjust optical power until you reach a bit-error rate of 10^-12 and perform layer-2 tests.

Layer-2 tests include Ethernet protocol analysis and tests of a network's OAM capabilities. Eric Lynskey, director of the EFM Consortium at the University of New Hampshire's Interoperability Lab (www.iol.unh.edu), says that the lab's compliance testing includes adding physical and frame errors to a link and monitoring how well a product performs link monitoring and remote-failure indication. Typical errors include errored frames and packet jitter at layer 2. At layer 1, tests include BER in the presence of noise, bit jitter, and low signal power (Ref. 4).

The lab uses much of the same equipment—sampling oscilloscopes, optical power meters, optical and copper spectrum analyzers, pattern generators, and protocol analyzers—that it uses to test other Ethernet technologies. As of press time, EFM-specific test suites are available only from Iometrix, but others will surely appear on the market.

You also need to perform the higher-layer tests to verify that your equipment meets customer expectations. For example, you'll need a protocol analyzer to decode higher-layer protocols. In particular, you need to decode and analyze at layer 3 (Internet protocol) and layer 4 (transmission-control protocol), says Mark Fishburn, VP of technical strategy at Spirent Communications. Fishburn notes that packet loss at layer 2 can create errors at the upper layers.

For example, data errors at layer 2 can result in errors in IP packets that route data to switches and routers. Thus, IP errors can cause misrouting of data among switches and routers. TCP provides checksums, sequencing information, and flow control, so a layer-2 error can produce unreliable TCP connections between network nodes.

Finally, you'll have to perform protocol analysis and functional testing at layer 7 just as you must for any network. For example, you'll still need a full suite of voice over Internet protocol (VoIP) tests such as voice quality, regardless of the transmission medium and underlying protocols.