Trading off power consumption and delay in packet forwarding engines with adjustable service rate

This paper presents an Adaptive Rate (AR) control policy for trading off power consumption and performance of packet forwarding engines. We assume a Forwarding Engine (FE) whose power can be scaled through Dynamic Voltage and Frequency Scaling (DVFS) and Clock Gating (CG). In a typical architecture, data packets are always transferred at the maximum speed consuming the maximum power. We proposed to tradeoff energy efficiency and forwarding performance of the FE by acting on the power-delay product. To this purpose, we provide an optimization framework for dynamically adjusting the capacity of single server queuing system. Then, we design a practical AR policy, which can be easily extended for multi-core/multi-processor architectures. The performance evaluation revealed that a significant power reduction is experienced when using our policy, with a small increase of packet delays during low traffic condition and without compromising performance during peak hours.