By Casimer DeCusatis
Today was the busiest day yet at OFC, with a full schedule packed with plenaries, invited speakers, tutorials, and much more. I’ve barely had time to write it all up, but let’s take a quick look at some of the day’s highlights.
The OFC plenary sessions features an interesting talk from the MIT Media Lab on the many ways that algorithms shape networks and the people who use them. Most people take pervasive network connectivity for granted; whether you’re using your cell phone to map out the nearest coffee shop, or just to navigate around the conference with the OFC app, portable networked devices have become an intimate part of our lives. But at its core, a smart phone is basically a collection of algorithms making autonomous decisions that many humans would find difficult. And these algorithms surround us in other ways, from the environmental controls in our buildings to the systems that run the elevators in our skyscrapers. And that’s not even counting the stock market, Google, or the emerging Internet of Things. The number of autonomous algorithms that affect our daily lives is truly astonishing, and will only increase as we continue to automate networking and cloud computing functions. This talk presented some unique examples of how algorithms affect our lives, from so-called “micro-crashes” on the stock market to the clever folks who developed a way to thwart traffic cameras using an SQL injection attack on their license plates . Those of us who create those algorithms for a living need to be aware of this complexity, and how our efforts will shape the next generation of networked devices.
Another OFC plenary from the CTO at Juniper dealt with the future of transport networks as a combination of software defined networking, network function virtualization, and physical device automation. According to this presentation, the “natural order” of networking involves using photons for networking (being bosons, they interact weakly with each other and are well suited to this task) and electronics, specifically routers with integrated optical ports, as switching nodes (since electrons are fermions, they interact strongly with each other and make the best building blocks for computing and routing). Both optical bandwidth-distance products and electronic transistor-Hz products have grown by a factor of 10,000,000 in the past few decades, which is driving the exponential growth and value of the Internet. The interaction of fundamental networking technologies such as packet switching, routing, and IP for scalability over the next five years will determine a great deal about how optical networks evolve to serve the needs of the industry.
Similar viewpoints were echoed by transport providers including NTT, Orange, and Huawei in their panel on technology and standards for the optical access layer. Specifically, the access network has received a lot of attention recently, and has been called upon to perform new roles using technologies such as passive optical networks (PONs). As the standardization for NG-PON2 nears completion, and next generation efforts including EPON begin to define their approach, the passive network is now being used to backhaul traffic from other high speed networks. Will wireless front-haul networks provide additional drivers for PON development, or at least for distributed, hybrid copper-optical systems? While it’s always difficult to predict the future, these trends seem poised to change the scope of access networking in significant ways.
Flexible, agile, adaptive, dynamic…these are all characteristics of emerging optical networks, at every scale from within data centers to metropolitan and wide area networks. During the panel on wavelength tunable and programmable optical transceivers, these characteristics were applied to all manner of optical network devices. Today in the transport network, reconfigurable optical add/drop modules (ROADMs) which can be adjusted to flexible wavelength grids have become standardized. Optical components which are bit-rate adaptive, with programmable modulation techniques, are starting to emerge as part of an elastic optical network. Many component technologies are being developed to provide a higher degree of flexibility in the physical layer, and leading edge service providers are beginning to deploy services which take full advantage of this bandwidth flexibility.
As the day wound down, there was still time for one of OFC’s famous Rump Sessions, an informal discussion of pressing questions that affect the industry. This year’s session focused on the shifting emphasis between hardware and software in the network. Some industry trends tend to favor hardware commoditization and software control of the network, including SDN and NFV. This has many benefits, for example software monitoring enables administrators to bypass failing optical links and provide geographically diverse back-up without expensive, high reliability optical components. There are other trends which suggest custom hardware is not likely to remain a viable investment option for much longer. Carrier revenues are basically flat, while bandwidth demand increases exponentially; this environment isn’t well suited to custom hardware. Similarly, cloud environments are growing rapidly, but their business models are mostly based on inexpensive, commodity hardware. The ROI for advanced optics isn’t very attractive (in particular, the high initial cost of 100G links has been criticized), due to the expectations that the cost of advanced optical hardware should be equal to or less than prior generations, even when volumes are low. Running counter to this trend, the development of integrated silicon photonics holds great promise for new hardware innovation, and the deployment of disaggregated network architectures. The promise of a fab-less design model, similar to that used by ASIC designs in the integrated circuit industry, may hold potential for shorter, low cost design cycles for optical hardware, and create new opportunities for innovation. Optical switching applications within cloud data centers may represent another new frontier for optical hardware. The tradeoffs between these two viewpoints always leads to a lively discussion among OFC attendees.
Do you think hardware innovation is dead, or is it poised for a comeback? Drop me a line on Twitter (@Dr_Casimer) #OFC and let me know…or if it takes more than 140 characters to express your opinion, I’ll be at OFC the rest of the week to sit down and chat. And don’t forget to check out my tweets of the most interesting new product developments from the OFC exhibition floor.
Posted: 25 March 2015 by
Casimer DeCusatis
| with 0 comments