Parallel processing in computer systems is gaining importance: not only supercomputers, but today also desktop and laptop computers have multiple processors. To make all these processors work efficiently on a single problem, they have to communicate. To this end, processors are connected to each other, and to the outside world, through a communication network. This network should allow a fast exchange of information, such that the processors receive new data to process in a timely manner. Current technologies, using electrical signaling, are reaching the end of their capabilities, however. Moreover, the traffic on such networks is very irregular, making some parts of the network highly saturated while other parts are barely used. The precise patterns also depend on the application, and can even change during the run time of a single application. This makes it very difficult to build an efficient communication network. This doctoral dissertation explores the possibilities of optical, reconfigurable networks. Optical connections are one part of the solution to the communication problem, since they allow for much higher data rates. Moreover, optics allows the network to be reconfigured, during the course of a program, in a data-transparent way, to the current traffic patterns. This way, all network connections are utilized more efficiently, which can both increase the network`s performance and reduce power usage.