Gigabit Ethernet is key weapon in telecoms standards battle

An ever-increasing demand for more bandwidth within the telecoms industry is paving the way for major opportunities for a new generation of companies able to provide state-of-the-art solutions. But can the traditional firms fight back by making futher use of their existing networks? Paul Gannon reports.

Telecoms technology is presently in a state of constant turmoil. The latest topic causing great debate is the role of agigabit Ethernet'. This asouped-up' version of the dominant Ethernet local area network (LAN) data transmission standard is currently used within enterprise LANs, for example, to link high-capacity servers or to provide services such as video-streaming direct to desktops.

But, whereas the basic Ethernet standard operates at 10Mbit/s, the gigabit version transmits data at one hundred times that speed ­ at 1.025Gbit/s. A medium version, at 100Mbit/s, is also available so users can deploy the appropriate speed Ethernet cards and devices to make the most of their LAN. This means that Ethernet is inherently flexible and cost-effective as a data transmission technology, operating at layer 2 of the stack of protocols to provide framing and transport functions between LAN nodes.

The problems only come when users want to link two LANs. Here they hit the traditional bandwidth bottleneck at the edge of the wide area network (WAN). Users have had to scale down even the lowest speed Ethernet transmissions to 2Mbit/s or less using leased lines or managed WAN data services such as frame relay.

The cost of wide-area bandwidth is so high that many users have been restricted to very low speed services ­ often as low as 64Kbit/s, the same amount of bandwidth as required by a single telephone call within the traditional voice network.

This sharp separation between the LAN and WAN has been so much part of the telecoms background scenery that equipment vendors as well as users have been trained to expect an expensive disruption at the boundary between user premises and the telco networks. No one expected a standard such as gigabit Ethernet to extend beyond the LAN so its performance parameters were restricted to short distances ­ a couple of kilometres at 10 or 100Mbit/s, and down to a few hundred metres at gigabit speeds. Restrictions apply to both copper wires and optical fibres.

So, if you wanted to push gigabit Ethernet into the WAN, you could not buy gear that would allow you to do it. And even if that had not been the case, it would have been mightily inefficient ­ the common telco wide-area layer 2 transport standard, SDH (synchronous digital hierarchy), offers only a few rigidly constructed transmissions speeds ­ 51Mbit/s, 155Mbit/s, 622Mit/s, and 2.5Gbps. None is efficient at carrying a 1.025Gbit/s gigabit Ethernet transmission.

Of course, all this suits the incumbent telcos, and quite a few of the new entrants as well. Not only do users have to pay high prices for broadband connections, but they also have to take on board the need to use wide-area protocols, such as frame relay or ATM (asynchronous transfer mode). The cost of staff skilled in these arcane technologies adds to the expense of using wide area communications and makes it attractive to use the telcos' high-margin managed services.

New development

But all that is now changing. Over the next few years Ethernet, in its low and high speed versions, will increasingly extend out into the WAN ­ raising the prospect that the distinction between the two will slowly be eroded away. The factors behind this new development are the onward march of IP ­ the Internet Protocol ­ and the revolution in optical technologies.

It is quite possible to take a gigabit Ethernet signal and put it through some devices that regenerate the signal at telco WAN standards for onward transmission over a fibre network. Obviously timing and co-ordination are important and so far this service has been possible only at the cusp between the LAN and the WAN ­ the metropolitan area network (MAN) within a single city for example ­ where a single fibre pair can be allocated to a dedicated path between sites.

However, this has proved to be an attractive proposition to users. As a result, gigabit Ethernet is gaining ground over alternative computer communications protocols such as Fibre Channel and ESCON also used in MANs. But now this optical service is creeping into wider areas. One company, Storm Telecommunications, even offers international gigabit Ethernet connections between nodes on its pan-European network.

"We claim to be the first company to offer international gigabit Ethernet services,“ said Steve Sanderson, chief executive of Storm. "Gigabit Ethernet is widely deployed at the desktop and in LAN backbones and now users can transport it internationally.

"As IP becomes the technology of choice, international gigabit Ethernet offers an end-to-end service with no need to convert to SDH.

"While the overall price-per-packet of the service will be broadly similar, it offers much lower cost implementation. Router technology costs about $1000 per gigabit Ethernet port, but it will cost 20 times as much for SDH interfaces, so gigabit Ethernet is inherently cheaper.“

New Generation Networks

Meanwhile, the adoption by equipment vendors, telcos, service providers and end-users of IP as the common data communications standard has led to work to strip out intervening protocols ­ such as ATM and SDH ­ so that IP can be carried directly over optical signals between routers. The aim is to develop anew generation network' (NGN) architectures that provide a common transport platform for a variety of services from a variety of providers using atransparent' optical channels.

The key to the process is the surge in the capacity and capabilities within the optical layer of the network. For example, SDH operates at layer 2 and currently provides aself-recovery' functions so that a break in one transmission link sees traffic on that link seamlessly switched to another link within 50 milliseconds or so. Data programs, or voice callers, at either end of the link will not notice the service break.

It should be possible to handle this function within the optical layer as vendors add aintelligence', in the form of software, to manage the network of optical fibres. Shifting this type of function out of layer 2 means that SDH could be reduced to simply aframing' data for transport between two nodes.

SDH then starts to offer no advantages over any other layer 2 framing protocol ­ such as Ethernet ­ and may indeed bring in disadvantages due to the inflexibility of the SDH transmission module speeds and the equipment needed to upgrade SDH links to very high speeds.

Over the past decade vendors have introduced massive capacity upgrades into optical fibre. A few years ago most telcos would have considered a single 2.5Gbit/s signal in a fibre as over-provisioning. Today it is possible to push up to 160 signals at 2.5 or even 10Gbit/s into a single fibre.

Within a few years we can reasonably expect to see wave division multiplexing (WDM) system that will support thousands of different wavelengths or optical channels in a single fibre, each carrying 10Gbit/s or more of traffic.

As telcos deploy these new systems they find that they can offer their customers bandwidth in unheard of capacities. In traditional terms 2Mbit/s is a big fat pipe; but with WDM telcos find it practicable to fulfill user demand, from ISPs for example, for capacities such as 2.5Gbit/s.

Clearly, it depends on the price of high-capacity optical bandwidth relative to traditional leased-line prices, but it soon becomes economic to start thinking about using a 2.5Gbit/s optical channel to carry gigabit Ethernet. The prize for the user is the erosion of the distinction between the LAN and the WAN.

"The ever-increasing demand for bandwidth is taxing the traditional telecom infrastructure and creating new opportunities for service providers with unique approaches to building next-generation service platforms,“ said Sten Nordell, chief technology officer at Utfors, another telco using advanced optical technologies to provide long-distance gigabit Ethernet services.

"Optimized for the IP and data-centric applications driving network growth, innovative new network architectures are leveraging state-of-the-art technology advances in areas such as IP, gigabit Ethernet and optical networking, and are paving the way for a new generation of IP-based broadband services.“

But all this will be very complex. Current optical technologies can handle recovery of whole fibres, but are some way from practical systems for switching thousands of different transmissions on individual wavelengths across a meshed core network between end nodes ­ which is what will be needed for widespread use of gigabit Ethernet in the WAN.

Intelligent nodes

The core of the anew generation network' will consist of intelligent software nodes that can switch wavelengths across the mesh of optical fibres. Each wavelength could be used to carry multiple IP traffic transmissions from different end-users between big routers ­ or could carry a single user's traffic using a protocol such as gigabit Ethernet.

If new entrant carriers are able to construct practical, robust networks, they could really upset the incumbents' data revenues ­ just as their voice revenues are also increasingly at threat too.

But the incumbents may be able to fight back. Given the competitive pressure, telcos are asking vendors to provide ways of carrying gigabit Ethernet in SDH modules. If that happens the incumbents could use the extensive reach of their existing SDH-based networks to beat off the new entrants ­ although they will never be able to match the costs of optical solutions.

Whatever happens, gigabit Ethernet looks like turning out to be a key weapon in the telecoms liberalisation battle.

Recent Issues