Abstract
Computing Systems are exponentially increasing their dependence on optical interconnects to meet their scaling bandwidth needs. Steady increases in computation density have put pressure on the interconnect infrastructure to keep up. The requirements for these interconnects include a critical set of metrics, that historically have focused on cost, but increasingly consider power and density. Furthermore, reliability (component and data) and latency must be considered as well. Thus the path forward to increasing bandwidth in these systems is becoming an increasingly complex set of trade-offs. This abstract describes the major applications for optical interconnects in computer systems, the relative metrics for these applications and considers new technologies in this context. © 2012 IEEE.