At conferences and trade shows there’s often a discussion around the value of Reconfigurable Optical Add/Drop Multiplexers (ROADMs). The idea of these devices dates back to the mid-90s when the DWDM industry was born. Now there are tens of thousands of such devices deployed in DWDM networks worldwide.
When ROADMs were first proposed, network architects predicted a rapid increase in the number of waves that DWDM could support on a fiber – a prediction that soon came true. A ROADM was designed to be able to add or drop one or more waves onto a fiber while adding the minimum possible cost to waves that are simply passing through this node.
The general idea was to get away from manual fiber patching to avoid truck rolls to intermediate nodes. ROADMs were designed to work as all-optical devices – basically mirrors and prisms (although the underlying technology is somewhat more complex than that!) – and they were supposed to be cheap to buy and simple to use.
But when service providers began to deploy the first generation of ROADMs, they hit a problem. The cheap and simple designs did not allow them to drop an arbitrary wavelength color at any given port. If they wanted to add/drop a wave, network planners would have to decide that ahead of time and choose the right wavelength for the transponders they ordered. This was difficult, obviously; this kind of ROADM limitation put a lot of pressure on traffic forecasting and reducing manual reconfigurations.
ROADM manufacturers responded – and produced “colorless” ROADM designs. While adding some expense and internal complexity, they almost eliminated the color limitations for add/drop. This was a step in the right direction.
But by this time ROADM networks were growing in size, and inevitably there would be more interconnection – or meshing in the network. Purely colorless ROADM designs were not that common because there was a need to support multi-way ROADM junctions. In these topologies it turned out that a given wavelength might not be able to be redirected to an arbitrary egress port – or direction of the ROADM. So the ROADM designers came up with colorless, directionless (CD) ROADMs. While adding more expense, internal complexity and internal optical loss, this was another step in the right direction.
The most recent planning problem is caused by the fact that the number of wavelengths on a fiber is constrained – typically 80 wavelengths in the C-band (although Infinera’s Double Density capability offers 160 wavelengths). This constraint means that it’s possible to have wavelength contention in a busy ROADM – where the wavelength being used for a particular service is already in use where you’d like the traffic to go. So the next “goal” is to get ROADMs to a place where they deliver on the original promise – contentionless. To produce a colorless, directionless and contentionless ROADM (CDC ROADM) would require a significant increase in complexity of the ROADM design. But right now there is no easy option for all-optical CDC, and the Optical Electrical Optical (OEO) method of wavelength conversion is perceived to be expensive.
But wait…there is a way to do it! As Mike pointed out in his blog last week, OEO doesn’t need to be expensive. By using large-scale photonic integrated circuits (PICs) we can shrink the optics down into a chip, and all those wavelengths become available in a small space – so we can co-locate electronic switching in what we believe to be a much more cost-effective form factor. And in fact Infinera has been shipping a CDC ROADM since 2004 using tried-and-tested electronic switching.
This approach of using PICs integrated with OTN switching delivers cost-effective CDC ROADM functionality for wavelength switching, but it also has another high value benefit – providing granular grooming and service delivery from ODU1 (ODU0 planned for the future) – which enables fast service delivery and highly efficient use of large DWDM pipes. This is the key to the Digital Optical Network architecture – an architecture that delivers operational simplicity, rapid rollout and more profitability to our customers.
ROADMs have always been a good idea and the improvement of colorless and directionless were the right steps forward. Contentionless is what is really needed next, and I contend that the cost-effective PIC-based OEO is the best approach to achieving this in the short term.
We believe that all-optical ROADMs will continue to be a useful tool in the network designer’s arsenal. Just as an all-optical approach will not solve all network design problems, so an all-OEO approach will not either. As bandwidth demands continue to press service providers to build bigger networks without increasing the size of operations teams, it’s anticipated that we’ll see that a combination of optical switching and electronic switching will likely be the key to the next generation of transport network bandwidth management.