The New Optical Internet
By Dr Joe Parker
It’s hard to keep a good idea down. The concept for an All Optical Network, or as it is now frequently called an Agile Optical Network or Optical Internet, is not a new one. Over ten years ago, testbeds such as MONET, AON, TBONE and others were testing the limits of optical networks, both physical and logical. Just-In-Time (JIT) switching, GMPLS and other protocols were designed and tested. Over the past 5 years many companies were built on the idea that exploding bandwidth demands would soon force more and more optics into the network.
The benefits of eliminating OEO conversions were obvious:
- Lower costs
- Lower heat and power
- Protocol transparency
- Higher efficiency
A new world was inevitable. It didn’t happen. Real world needs never met our expectations. Most of the companies died and the nuclear winter began.
Spring is coming. Every day we hear calls for a more agile optical network from both carrier representatives and industry pundits. Conferences are starting dedicated to control planes for photonic networks. But why? What is so different today from 5 years ago? The benefits haven’t changed much. Whether it be ROADMs or photonic crossconnects, tens of watts are always better than kilowatts and a simple upgrade from 10Gb/s to 40Gb/s is always better than a forklift. True, optical devices have improved their performance and reliability, but not enough to cause a dramatic shift in optics’ fortunes. The failure 5 years ago was a failure of demand and it is to demand we must look to discover the cause of the new resurgence.
Bandwidth demand is surging. Multiple sources show broadband traffic doubling every 18 months. This demand is driven both by the dramatic increase in broadband subscribers and the flood of new telecom and Internet-based services. But, many of these same forecasters were reporting astronomical Internet growth 5 years ago. What has changed?
In retrospect, the old analyses were weak. Based almost purely on the burgeoning numbers of Internet users, they were insufficiently grounded in the drivers needed to continue the growth. The current analyses are much more thorough in their study of the realistic adoption rate of high bandwidth connections and the traffic necessary to fill the pipes and supply a profit stream for the carriers.
It is the nature of the emerging video content of all flavors, VOD, IPTV, ITV, etc., which not only lead credance to the forecasts but also provide a major driver for a New Optical Internet. The world is rapidly becoming IP/Ethernet. The improved economics, both CAPEX and OPEX, have won, as usual. The improved economics, however, do not extend to streaming media. IP and streaming video do not play well together. IP is, at its core, a best effort protocol. Its reaction to congestion is to drop packets and use retransmission to recover. But, streaming video requires a high QOS channel, i.e. low packet arrival time variation and no dropped packets. There are only two ways an IP network can achieve this.
First, massive over-provisioning. Keeping the network utilization under 15% minimizes video degradation. However, network utilizations this low are economically unacceptable. Second, minimizing the number of IP hops. Degradation is directly proportional to the number of hops. A hybrid network consisting of IP routers/switches and photonic ROADMs and OXCs will permit an intelligent edge to direct high priority/high value traffic to the higher quality, pure optical paths.
After a decade, we are now seeing the resurgence of optics starting with ROADMS and continuing with photonic OXCs, a resurgence not driven by optics device innovation, but by the needs of the users. It’s about time.
