Abstract
Deep packet inspection is perceived as an energy-hungry technology, but in this paper we demonstrate that the knowledge it provides can lead next-generation green Internet technologies to save large amounts of operating power on processors and links. Green technologies scale the power consumption of links and devices in proportion to the workload to be processed and transported. Deep packet inspection can classify flows within this workload in latency-critical and throughput-critical. We show through a novel theoretical analysis for Markov-modulated traffic models that the processing of common throughput-critical flows demands about half the power that the processing of their latency-critical counterparts does. Subsequently we introduce a novel simulator of systems composed of advanced green processing cores and Energy-Efficient Ethernet (IEEE 802.3az) links. On this simulator we present the evaluation of energy budgets for real-world traffic and confirm our theoretical work, demonstrating a power saving of more than 40 % on green processors and links.