Avoiding the metro Ethernet brick wall.

"Ethernet always wins" is possibly a little-known quote in telecoms, but familiar to many in the Enterprise networking space. Despite a history of arguments against it, Ethernet has seen many technologies, some arguably superior, join Betamax on the scrap heap. In nearly 20 years in high technology, I have never seen any mud thrown at Ethernet actually stick to it and prevent its spread into nearly every area of life. This is becoming reality for all of us; anyone who has recently experienced the joys of a working broadband connection following years on dialup has probably noted that the connection from their cable or DSL modem is Ethernet.

The basic reason that Ethernet has been so successful is that it works, is ubiquitous, and most critically, is adaptable. This in turn leads to Ethernet and variations on it being by far the cheapest practical way to solve data communications problems. There are always more technically elegant ways to do the job, but the sheer costs of development and equipment never seems to prevail over Ethernet's chameleon-like ability to adapt itself to the challenge of being good enough to facilitate rapid, cost-effective deployment and revenue generation.

The next challenge in the crosshairs for Ethernet is moving beyond the Enterprise network to the connections between them--in the majority of cases, over public networks.

The reason many enterprises are interested in Ethernet LAN interconnection is simple. Ethernet has essentially won as the network of choice within corporate networks, and in different offices that need to be linked. The historical model for branch interconnection has evolved from transaction-based enterprise applications centered around mainframe programs accessed by emulators such as 3270 or 5250 over low-speed Frame Relay networks. Newer Enterprise applications are being developed on distributed platforms built around peer-to-peer architectures such as Java and Microsoft's .Net. This and the rise of interdepartmental e-mail culture are straining the bandwidth available from historical network architectures.

A typical telecom network offering is based around traffic management of bandwidth over its TDM network (either SONET or SDH). This network is uniquely unqualified to carry the newer types of bursty LAN traffic, and so traffic engineering becomes an expensive problem. Enterprises, on the other hand, after decades of throwing bandwidth at local LAN problems, have had to rein in their natural tendencies due to the high cost of high-speed TDM services beyond T1 or E1.

The combination of application-driven customer demand factors and telecom network architecture is being targeted with the enormous capital returns generated by Ethernet in the enterprise space. These dollars manifest either as venture capital, or as company equity value, and are being targeted at Ethernet-based evolution of Telecom networks--both for data and voice traffic.

As telecom network architectures evolve, they run slap into the biggest challenge to broadband services of any kind: how to get the required bandwidth over the last part of the network to their customers. The standard answer to this challenge has always been to deploy fiber. And why not? Fiber offers infinite capacity and is immune to electrical interference. The answer is that even investing the $10B of cash flow generated annually by all US RBOCs (source: Deutsche Bank) it would take over 30 years to provide a connection to every business, let alone every residential, subscriber.

It is quite fashionable to repeat stories of subscribers who receive dirt-cheap Ethernet from some new provider, but the practical reality is that you have to be one of the lucky few to be "on net" otherwise you get nothing at all.

So if there is such a demand for Ethernet, how is it to be addressed? One solution is to give the user a box with an Ethernet interface and a slow network connection (a bit like the bad old days of dialup, having a 56K modem connection actually delivering 4K of traffic), this may be the preferred approach to residential broadband, however it is not acceptable to business, which require guaranteed symmetrical bandwidth. An alternative approach is to take the traffic capabilities of single copper pairs and aggregate them to create a connection that works at a recognizable and acceptable speed all the time. Using the existing copper network in this way overcomes the capital investment problems of fiber, yet offers native Ethernet connectivity at Ethernet speeds. On this basic physical layer, the telecom provider is then free to build added value service layers such as VLANs and MPLS capabilities.

The first attempts at Ethernet delivery using more than one circuit included Multilink Frame Relay and MLPP, by far the most popular has been Inverse Multiplex ATM (IMA) bonding ATM circuits out of a standard DSLAM. The problem with all these solutions is that they are overlaying Ethernet on top of the more expensive architecture that Ethernet is designed to replace. This creates problems for deployment, management and is an inefficient use of network bandwidth.

A more fundamental problem with these approaches is the loss of flexibility in traffic type afforded by these compromise options, in order to deliver either the quality or the bandwidth, certain key characteristics of delay and error-sensitive traffic are sacrificed, compromising the Ethernet's link to be a substitute network connection for all types of contemporary LAN traffic, including the more priority-sensitive traffic such as Voice over IP and real time video.

Solutions to this problem come in two forms, currently the subject of several standards body efforts. The first is to define a high-speed standard for Ethernet over short copper loops, using VDSL technology, which is designed to be capable of symmetrical service. The second, and more viable, for the longer loops found in the carrier serving areas (CSA) of facilities-based carriers is the use of symmetrical g.shdsl modems. These two approaches form the short and long reach physical layer definitions respectively of the 802.3ah Ethernet in the First Mile (EFM) standard.

Taking the EFM standard and applying advances in link aggregation technology, it is now possible to provide multiple megabits of symmetrical Ethernet bandwidths over extended distances, up to 18,000 feet, far beyond the typical Carrier Serving Area (CSA) limit of 12,000 feet (both these distances on 24 AWG cable). Products shipping today such as the MetaLIGHT 100 from Actelis Networks have proven their use in extending the fiber-based Ethernet services of carriers out to copper-fed customers with no compromise on service quality, or traffic delay or reliability. Such standards-compliant products stand to transform the economics of Ethernet availability to the 89% of businesses that today have no fiber connection.

With the capex barrier to service deployment removed, carriers and end users alike are free to enjoy the cost and efficiency benefits of ubiquitous Ethernet networks. Once again, Ethernet wins.

www.actelis.com

David Perry is area vice president, corporate marketing, at Actelis Networks (Fremont, CA)