Time-Encoding Scheduling

This paper develops a novel scheduling mechanism for packet switches based on a minimal set of hardware components. The standard model of computation implemented by digital logic is replaced by a model based on time encoding. This new model allows for distributed computation with low hardware complexity. A distributed switch scheduling algorithm utilizing time encoding is shown to deliver performance comparable to centralized algorithms under uniform traffic. Exploiting a connection between switch scheduling and interval packing, we argue that the distributed nature of the algorithm limits the maximum relative load to 1 − e under diagonal traffic. The stability of the algorithm can be improved by enabling reversibility in distributed decision making.