Convergence and Sensibility

By Fujitsu's Sam Lisle

Convergence is perennially hot.

You know ... convergence. Convergence over MPLS, convergence over DWDM, Ethernet as a convergence technology, etc. Everywhere I’ve been for the last 20 years, people have been talking about convergence of one kind or another. Some of these convergence ideas have been great. Some have been tragic.

For example, the worst pitch I ever heard occurred at the height of the bubble where a start-up CEO talked at a conference about his company’s industry-shattering development – a single box that was a SONET multiplexer, an ATM switch, a frame relay switch, and an IP router all in a single package. Even worse was that the audience sat in rapt attention and appeared to buy every word. (Perhaps they were VCs who literally were buying words?)

In this obvious case of “convergence gone wild” the platform and company met with ignominious failure. Invariably, these types of massive god box convergence platforms (that was not the only one) never really make it in the marketplace. Creative but over-ambitious technology guys looking for a differentiator glom together as many functions as they can possibly find, and the result is a contraption that is not really very good at any one of the included functions and always gets clobbered by a platform that excels at solving a meaningful problem.

There have been, however, many examples of successful technology marriages. In the transport world, ROADMs have successfully integrated DWDM and SONET ADM functionality. In the world of service elements, standalone ATM and frame relay switches gradually gave way to multiservice switches that could switch and interwork both kinds of traffic, and there have been successful examples of integrated Ethernet bridges and IP routers.

So what kinds of technology convergences work? Are there boundaries that sensible convergence operates within and “convergence gone wild” gleefully ignores? One clear boundary that has always existed in telecommunications has been the line between transport/aggregation equipment and service equipment. This boundary is a reflection of both fundamental equipment technology differences and deployment and operational differences.

Transport/aggregation equipment has always been specialized to gather a bunch of traffic together (aggregation), ship it over a distance (transport) and protect it along the way. Transport elements, like SONET ADMs, have been historically inexpensive because they were a cost that underpinned every service and simple to manage because they were highly geographically distributed. Service elements like ATM switches and IP routers have contained the specialized intelligence and processing capability that is unique to the actual service being sold, then have handed the processed traffic to the transport/aggregation environment for efficient carriage downstream. Service elements have historically been more costly per bit, but the cost was justified because the platform existed to deliver a specific service. These elements have been more complex to manage, but since there were fewer of them and they were centrally located, that complexity was bounded.

The sensible technology convergences have always been isolated to either one side of this boundary or to the other, and every time (at least in my memory) a convergence effort attempts to go wild and cross this boundary, the effort flops.

The latest convergence buzz focuses around the issue of providing integrated aggregation and transport for a distributed metro network dominated by Ethernet and IP. This is a great opportunity for convergence because aggregation and transport functions have historically been integrated into single platforms.

When networks were dominated by DS1 and DS3s, SONET elements provided converged aggregation and transport in the same package. As networks began to evolve toward Ethernet and IP, these functions have temporarily diverged and this divergence has been rough on economics and scalability. Transport in Ethernet/IP networks is provided by ROADM/DWDM systems while aggregation has been handled in a variety of ways by disparate pieces of equipment. Sometimes the aggregation function is handled by service elements like IP routers and Ethernet bridges bringing additional cost and complexity to the network. Sometimes Layer 1 SONET elements deliver the aggregation with limited granularity and flexibility. Sometimes the aggregation function is not performed at all, resulting in inefficiency.

Providing the economical, reconvergence of aggregation and transport for Ethernet and IP traffic is exactly the problem that packet optical networking solves. This new class of solution marries DWDM transport to connection-oriented Ethernet and Layer 1 aggregation. Many vendors have been rolling out converged equipment with exactly this type of focus and service providers are beginning to evaluate and deploy this new equipment type.

This latest convergence uproar is still fairly new, and even among sensible convergence approaches there are important differences. As these differences become better understood and as the connection-oriented Ethernet protocol wars die down (more on this topic another day), the dust will settle on this convergence discussion and we can all get on to the next one. Meanwhile, the important problem of Ethernet/IP aggregation/transport and the emerging set of packet optical networking solutions is guaranteed to keep convergence a hot topic for a while longer.

Sam Lisle is market development director at Fujitsu Network Communications. He can be reached at sam.lisle@us.fujitsu.com.

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines