A compressed and dynamic-range-based expression of timestamp and period for timestamp-based schedulers

Scheduling algorithms are implemented in high-speed switches to provision quality-of-service guarantees in both cell-based and packet-based networks. Being able to guarantee end-to-end delay and fairness, timestamp-based fair queuing algorithms have received much attention in the past few years. In timestamp-based fair queuing algorithms, the size of timestamp and period determines the supportable rates in terms of the range and accuracy. Furthermore, it also determines the scheduler's memory in terms of access bandwidth and storage space. An efficient expression can reduce the size of the timestamp and period without compromising the supportable rate range and the accuracy. In this paper, we propose a compressed and dynamic-range-based expression of the timestamp and period, which can be readily implemented in hardware for both high-speed packet-based and cell-based schedulers. As compared to fixed-point and floating-point number expression, when the size is fixed, the proposed expression has a better accuracy. Regarding to efficiency and relative error consistency, it even better than our earlier proposal.