We have accomplished the following key advances in each of these focus areas:

Fully interconnected 12-port Multi-Terabit Optical Packet Switched Network

Utilizing the paradigm of photonic switching elements which are transparent, broadband, and self-routing, we have constructed a complete 36-node 12x12 optical switch with Terabit routing capacity. This is the first complete implementation of a truly optical packet switched network. The system is capable of routing packets with 160 Gbps (10 Gbps x 16 WDM) payloads from any one of the 12 input ports to any one of the 12 output ports. The median (and average) latency for the system is approximately 110 ns, which corresponds to 5 node hops [Shacham 2005], [Small 2005], [Small 2005].

Investigations of the behavior of the network and its scalability [Small 2006] have demonstrated the power and timing flexibility [Small 2005], [Small 2005]. Recent experiments study the effect of injecting packets with variable numbers of payload wavelengths into the network simultaneously. This emulation verifies the feasibility of dynamically varying the payload size in-situ by altering the number of payload wavelengths encoded [Small 2006].

Nanophotonic Interconnection Networks

The miniaturization of switching elements afforded by the large-scale integration of nanophotonics raises a critical challenge to the internal processing of the optical data packets while maintaining a memory-free switching fabric. We have developed a new routing approach specifically designed to the unique nano-scale integrated interconnection network, termed SPINet (Scalable Photonic Interconnection Network) [Shacham 2005]. SPINet is a novel optical switching architecture that does not employ optical buffering of any kind and messages are dropped upon contention. A novel physical layer acknowledgement protocol is used to provide a drop-detection mechanism. Performance analysis demonstrating an average bandwidth exceeding 40 Gb/s per port has been reported and error free routing of 160 Gb/s peak bandwidth has been experimentally verified in a prototype switching node [Shacham 2005]. We have introduced the concept of path adjustments as a means of increasing the utilization of optical packet switched networks. Simulation results show substantial performance improvement under uniform and non-uniform traffic, and recent experimental demonstrations prove feasibility [Shacham 2006].

Silicon Photonic Enabling Network Elements

Recent advances in silicon photonic micro-fabrication techniques have led to the development of an optical toolbox consisting of the vital elements required for realizing chip-scale photonic networks. Collaborating with a few of the leading research groups in this area, we have developed a network-driven design approach for implementing and characterizing these building blocks. Among the components that we have investigated are an electro-optic form-factor translator [Lee 2007] that converts space-parallel data from an electronic bus into wavelength-parallel photonic data coinciding on a single waveguide link; a broadband, low-loss photonic link [Lee 2008] that can carry terabits-per-second of information encoded in this wavelength-parallel manner from one node to another; and a high-speed, multi-wavelength switch [Lee 2008], which enables dynamic message routing of the broadband signals in complex network architectures.

In addition to the design and implementation of these components, we have characterized their predominant signal-degrading impairments in an effort to further improve the performance of networks derived from future versions of these devices. Narrowband filtering of high-speed data signals is an important consideration when passing data through high-quality-factor resonators, as is done in the multi-wavelength switch. Therefore, we have experimentally and numerically characterized the power penalty induced by this effect [Lee 2006], and applied the numerical simulator to a wide variety of other higher order transfer functions, which are realizable with ring resonator architectures [Small 2006].

Characterization of the Physical Layer

The scalability and efficiency of optical packet switched (OPS) networks are critical performance parameters. The precise photonic implementation of the network physical layer may limit the physical size and usable bandwidth. Because SOAs are used as the key switching element in OPS networks, understanding their performance characteristics is important to optimizing the system and is a continuing theme of our research activities [Lu 2004], [Liboiron-Ladouceur 2006]. A re-circulating test-bed environment has been implemented to experimentally study the physical limitations of OPS networks. This test-bed has enabled unique physical layer analysis in the context of different modulation formats, including the first investigation of polarization mode dispersion in OPS networks [Liboiron-Ladouceur 2005], [Liboiron-Ladouceur 2006]. Modeling and simulation of the physical layer behavior are used in conjunction with the experiments to optimize the information carrying capacity and efficiency in OPS networks.

Scalable Optical Packet Buffers

In order to address the need for a practical solution for buffering optical packets, we have developed a novel optical packet buffer [Small 2007]. The buffering architecture is comprised of identical SOA-based building-block modules, yielding straightforward scalability and extensibility. In a time-slotted manner, the buffer supports independent read and write processes without packet rejection or misordering. Both first-in first-out (FIFO) and last-in first-out (LIFO) prioritization schemes have been realized. Further, active queue management (AQM) can be implemented on the buffer architecture to allow for network congestion control. Simulations have verified the improved buffer performance with respect to latency [Shacham 2007], and experiments have demonstrated the functional verification of the optical packet buffer modeling AQM [Wang 2007]. Modifications to the basic buffer architecture have also allowed for the experimental implementation of an optical network interface buffer, providing dynamic queue management and cross-layer signaling for an OPS network test-bed [Lai 2008].

Wideband Programmable Optical Routing Modules

We have successfully implemented multi-wavelength optical routing modules that included programmable functionalities and demonstrated the incorporation of priority scenarios directly into the optical layer [Shacham 2005], [Shacham 2005]. The controllable optical packet injection module mediates packet timing and insertion into the OPS switching fabric dynamically, on a packet-by-packet basis in the presence of traffic contention.

Optical Transparency and Ultra-low BER

Our paradigm for switching nodes and routing modules within OPS systems is based on a fully transparent design for the optically encoded information: the incoming optical data should match the outgoing optical data as closely as possible, in terms of optical power, optical signal-to-noise ratio, etc [Small 2005]. This level of transparency for implemented OPS switching node was demonstrated by determining the bit error rate (BER) introduced by the node itself. We introduced an improved sinusoidal interference method employed as a means of extrapolating ultra-low BER levels which are not ordinarily possible to obtain in conventional laboratory experiments [Small 2005].

Bit-Parallel WDM Message Exchange

In an attempt to explore the limits of minimal latency data exchange, we developed a single-bit DWDM bit-parallel transmission interconnection test-bed that demonstrates the feasibility of this optical link. The data payloads are entirely recovered and processed at the destination node using a novel embedded clock signal with a measured record clock-to-data skew tolerance window of 150 ps [Liboiron-Ladouceur 2006].