Scaling Terabit Networks

Overview

Read incubator
white paper.

This workshop brought together leading experts from academia, industry, and government to discuss the major technological obstacles and opportunities to efficiently scale networks in the Terabit regime. Focus areas include exploring how emerging network architectures drive new requirements for optical systems and how photonic components can help scale architectures to terabit capacities; rethinking the operating system interface and application interface to the network; exploring which optical layer characteristics to share and what higher layer functions can benefit; and, exploring optical and electronic networking cooperation and respective metrics.
View full agenda.

Goals

Bring focus to the major technological obstacles and opportunities for scaling in terabit networks. The goal of this two day, small group meeting was to develop Grand Challenge research initiatives by:

• Exploring new network architectures that drive new requirements for the optics and how emerging photonic components can help scale these architectures to terabit capacities.
• Rethinking the operating system interface and application interface to the network
• Exploring which optical layer characteristics to share and what higher layer functions can benefit.
• Exploring optical and electronic networking cooperation and respective metrics

Objectives

• Identify the technological challenges facing terabit networks and continued network scaling
• Establish key metrics, targets, and capabilities for efficient and scalable terabit networking
• Set research priorities and requirements to enable efficient and scalable terabit networking
• Identify application drivers and early adopters of this new technology.

Sponsorship

Support for this workshop is provided by the National Science Foundation. The Center for Integrated Access Networks, an NSF Engineering Research Center, and the Optical Society will co-host the workshop. Travel support will be provided for participants academic institutions upon request and as available.
 

Motivation

Commercial transmission and switching systems support 100 Gb/s, with edge rates routinely exceeding 10 Gb/s and efficient optical transport reaching to the enterprise and home. Terabit per second signal transmission has been demonstrated in labs, with commercial systems on the horizon. Emerging applications are expected to comprise bursty, large and unscheduled transactions, with resulting network traffic exhibiting heavy tail behavior. Component commoditization is changing the traditional balance of storage, processing and communications resources, making way for new architectures.  Furthermore, disaggregated computer platforms will bring optics past the top-of-rack switch directly to the end host. These trends demand research on innovative new terabit-scalable network architectures designed for end-to-end optical and electronic technologies.
 
New terabit-scale architectures must consider changes at all layers of the stack.  Optical devices are very different from their electronic counterparts, in some cases without an electronic analog, so it’s likely the optimum optical network architectures will not be the same as today’s Internet architecture. Scaling transmission to terabit rates has exceeded serial scaling methods, driving ever-more parallelism. Bridging parallel optical connectivity with multi-core processing will result in higher overall system utilization, but only if that connectivity can be used effectively.  Researchers must rethink how photonics can improve network scaling, and determine high-layer architecture changes that better exploit optics.
 
Today, Software Defined Networking focuses primarily at the electronic layer and is being explored for the transport layer. In this workshop we will consider optical communication and networking hardware technologies together will higher layer protocols for the exploration of new agile, programmable and efficient optical network architectures. Software programmable optical devices can figure prominently in new virtualizable and flexible architectures that incorporate optical layer agility and scalability. Cross-disciplinary research is needed for “programmable optics” to reach full potential and to address the concomitant control and management challenges for both systems and networks.
 
Networks and data centers that support scientific research are experiencing exponential traffic growth, and so research backbones (ESNet, I2, NLR) and regional networks are anticipated early adopters for terabit-scale networks.  On the commercial side, video production, digital healthcare, warehouse-scale-computing and latency-driven financial applications will benefit immediately from this new capability.
 

Program

The workshop objectives was addressed through plenary and breakout sessions over two days.  The first day focused on technological challenges, metrics, targets, and capabilities. The second day set research priorities and requirements (simulation, experiment, field trial, etc.). Research priorities will be ranked and evaluated relative to the metrics, targets, and capabilities. A workshop report was made publicly available through Optics and Photonics News.
 

Location

The workshop took place at OSA Headquarters in Washington, DC. OSA is located at 2010 Massachusetts Avenue, NW, Washington, D.C.