The Effect of Correlated Arrivals on the Performance of Energy Efficient Ethernet

Energy Efficient Ethernet (EEE), standardized in IEEE 802.3az, is designed to reduce energy consumption in local area networks by transitioning the network interface into a Low Power Idle (LPI) mode during idle periods. While this mechanism has been widely studied, most analyses assume packet arrivals follow a Poisson process or independent inter-arrival times, which may not reflect the bursty and correlated nature of real network traffic. In this paper, we model packet arrivals using a Markovian Arrival Process (MAP) to capture temporal dependencies and varying levels of correlation. We then evaluate how traffic correlation affects key EEE performance metrics, including idle time ratio and average packet delay, across different traffic load conditions. Our results reveal that correlation in packet arrivals can significantly impact the trade-off between energy savings and delay. The stronger correlation between inter-arrival times leads to a reduction in idle time ratio due to more frequent state transitions, while also causing an increase in packet delay, especially under low frame coalescing thresholds. These findings highlight the importance of accurate traffic modeling in evaluating the performance of EEE protocols.