IP and Optical: We’ve Come a Long Way, Baby
IP over DWDM
At several recent trade shows, most notably the NFV and Carrier SDN Event, Heavy Reading Analyst Sterling Perrin showed a slide demonstrating the industry’s long history with IP and Optical integration. One of the images on that slide is an article in LightReading from April, 2000 entitled “IP over Glass, Who Cares?”, a very interesting article to re-read from a historical perspective to see just how far the industry has progressed in 16 ½ years.
In the article, Ciena (then a relatively young WDM company) and Juniper (then a relatively young router company) announced interoperability at 10Gbps. Today, this type of interoperability seems trivial, but at the time, such a demonstration was controversial with “many experts” saying that it was not practical and claiming that the industry trend is “moving away from efforts to put IP directly on glass”. This statement could be enshrined alongside “everything that can be invented has been invented” and “I think there is a world market for maybe five computers” as one of the worst predictions in history.
What were the arguments against what now seems like an obvious implementation? The main argument was that once traffic was put onto a DWDM system, it could not be moved around flexibly. The standard approach in 2000 was to connect from a router to a SONET/SDH or ATM transport element, then to DWDM. The cutting edge alternative in 2000 was to put an “optical switch” in between the DWDM system and the router.
For those not in the industry long enough to remember, in 2000 an “optical switch” was typically an optical-electrical-optical (O/E/O) translator that took the traffic from a wavelength, converted it into the electrical domain, moved it around with a simple crossbar switch, and reassembled it on the other end with other signals onto a new WDM line. These O/E/O switches were all the rage for a few years, with companies competing to build larger and larger switching matrices and, consequently, larger and larger pieces of equipment.
Today, the invention and proliferation of ROADM technology has made the optical switch somewhat obsolete. There are still optical switches around that use optical technology such as mirrors to move light around, but they are largely relegated to lab and automatic patch panel applications. The flexibility and scalability of ROADMs, with their ability to break apart the multi-wavelength streams and route each wavelength individually have made the use of O/E/O “optical switches” in the network appear as a historical oddity.
Technologies such as OTN have taken over the sub-rate capabilities offered by SONET/SDH and ATM exists only in old textbooks and perhaps some remaining legacy DSLAMs that have yet to be replaced. Today, the term “IP over Glass” is rarely used, as this controversial topic from 2000 has become the standard implementation in 2016.
So what will be the obvious technology in 2032 that is currently being scoffed at by industry experts? Perhaps in 2032 we will look back on the beginnings of optical SDN in the same way we look at the IP/optical integration today – something that seemed controversial at the time, but seems completely obvious now.