Optical Reality Check

2:00 PM -- Over the past couple of days NASA Research and Engineering Network hosted an excellent workshop to develop a roadmap for next-generation optical networking. Representatives from many of the world's leading research optical testbeds and networks were present at the workshop and contributed to the discussion.

One of the best presentations was given by Bob Feuerstein of Level 3, who provided a well-argued reality check in terms of agile dynamic optical networks. Basically he stated that the reason customers buy optical network services is for reliability and security where they do their own routing and restoration at Layer 3. Protected wavelengths are poppycock, he claimed. He also stated that there is no demand for wavelengths on demand, and that all optical switching is hooey.

These comments run counter to many of the current optical research initiatives. In many ways current optical network research echoes much of the previous work done with ATM, such as bandwidth on demand (a.k.a. SVCs), QOS routing or scheduling, and reservation for high-bandwidth applications. These things never took off with ATM, and there is no evidence that they will be any more readily accepted with optical networks.

One of the fundamental assumptions of current optical network research is that will be lots of idle optical circuits available from which users can request bandwidth upon demand. But as Dr. Feurstein pointed out, the telco industry, like every other industry, is now only provisioning services just in time, as customers pay for them. The days of monopoly telephone companies with over-engineered network capacity to provide bandwidth on demand are long gone.

But we did learn some valuable lessons from the days of ATM – lessons which I think we can apply to the new optical networks. The two most important tools that came out of ATM are MPLS-TE and VPNs. In my opinion, the exciting research opportunities in optical networking therefore is building on what we learned with ATM and exploring next-generation traffic engineering tools and VPNs.

As many of you know, Canarie and our research partners around the world have been exploring these concepts with user-controlled lightpaths (UCLP) for user-controlled traffic engineering and next-generation VPNs – which we have labeled Articulated Private Networks (APNs).

To my mind, the low-hanging fruit for optical networks is in increasing demand by many organizations to own and operate their own wide-area network. So the more we can provide tools for them to manage these networks independent of any central network management organization the better. This will mean lower opex costs for the wavelength and fiber providers and new business opportunities to provide outsourcing to manage these customer-owned and -controlled networks.

A good example of this trend is the announcements made at the NASA workshop by many government agencies, such as USGS, NASA, DOE, and other large science departments to acquire their own wavelengths on the National LambadRail. Each agency will have its own private optical network – and now they need the tools to do their own add/drop, grooming, crossconnect, etc. independent of the other users on the network. This is exactly what UCLP is designed for.

The growth of the National LambdaRail to now include many agency networks is extremely exciting and I think genuinely represents the future of networking, optical and otherwise. NLR now has well over a dozen "condominium" customers on their network – all with their own individual wavelengths.

— Bill St. Arnaud is Senior Director of Advanced Networks at Canarie Inc.

Be the first to post a comment regarding this story.
Sign In