Traffic in today's backbone networks is dominated by Internet protocol (IP) applications and is continuously increasing. The increasing demand is driven by video streaming, cellular phone technologies, and the vast amounts of information stored in data centers. Traffic volumes exceeding 1PB (>1,000 Terabytes) per business day have been reported and are expected to increase by more than tenfold over the next decade. Current electronic core network routers rely on a store-and-forward technique that requires optical-electronic-optical (OEO) conversions of high-speed data in order to route packets. This technique requires each bit of incoming data to be processed by high-speed electronics which tend to be rather power intensive and do not scale efficiently with the increasing trends in traffic demand. Optical data routers (ODRs) and optical packet switching (OPS) are currently being investigated as potentially scalable, energy efficient alternatives to current electronic routers. The OPS scheme utilizes a packet format consisting of a low-speed header followed by high-speed payload. The routing of such a packet is performed by processing the header via low-speed electronics that are relatively inexpensive in terms of cost and power dissipation while allowing the payload to be transparently processed entirely in the optical domain. Adaptation layers are required at the edge of core optical networks to facilitate the interoperability between current electronic and future optical networks. Successful router demonstrations have relied heavily on fixed-length packet formats, which require edge adaptation layers to perform computationally complex fragmentation algorithms to break larger variable-length packets into smaller fixed-length packets. The fragmentation process is viewed as detrimental to performance because of its inherent latency penalties and by the fact that re-transmission of entire datasets is required if individual fragments are lost.

This dissertation presents enabling technology required for low-latency, low packet loss, interoperability between today's electronic networks and future optical networks. Adaptation between legacy 100 megabit Ethernet (MbE) and 40Gb/s OPS formats is successfully achieved by utilizing custom, FPGA-based Edge Adaptation Layers that demonstrate latencies below 300ns and packet loss rates less than 1E-5. Latency penalties are minimized by forgoing the fragmentation process at the edge and enforcing variable-length packet compatibility within optical core routers. A re-sizable optical buffer is designed and implemented to enable all-optical routers to accommodate packet lengths ranging from 40 to 800 bytes with less than 5% packet loss. This dissertation culminates in an end-to-end ODR link demonstration utilizing an energy efficient FPGA-based electronic control and UCSB-fabricated photonic integrated circuits (PICs) housed in custom package sub-mounts and FPGA-based driving circuitry. The end-to-end link demonstration successfully achieves adaptation between 100MbE and 40Gb/s optical packet formats, optical packet buffering, forwarding, and 3R regeneration (re-amplification, re-shaping, and re-timing) of 40Gb/s optical packets at packet loss rates below 1%.