Modeling and analysis of watchful sleep mode with different sleep period variation patterns in PON power management

Reducing power consumption in an access network has become an increasingly imperative design goal, due to the fact that information and telecommunication technologies contribute to an increasingly large proportion of greenhouse gas emissions. The passive optical network (PON) is considered as the most attractive and promising technology to provide low-cost services to end users in a power-efficient way due to the passive nature of remote nodes. However, it is necessary to further reduce the energy consumption of PONs, with the wide deployment of PONs and the rapid growth of data transfer rates in PONs. In this paper, we focus on the watchful sleep mode with different sleep period variation patterns for multiple optical network units (ONUs). This is because, in the traditional operational mode, ONUs have to continually listen to and inspect traffic from the OLT and hence always remain active even when there is no/light traffic, which contributes to the majority of energy wastage to the PON. We first modeled the watchful sleep mode based on the Markov chain model in order to analyze the effect of each key parameter on system performances in terms of the energy-saving efficiency and data packet delay. Due to the fact that the sleep state is the key to power-saving, we designed four different sleep period variation patterns (e.g., constant, linear_1, linear_2, and exponential patterns) to study the impact of these different patterns on the integrated performance in terms of the normalized cost value. Through extensive simulations, we found that, in the watchful sleep mode, the effect of the number (n) of (sleep, listen) state pairs would be insignificant, and the effects of other parameters on the performances are analyzed comprehensively. The minimum normalized cost value can be obtained under the four different sleep period variation patterns, which represent the optimal trade-off between the above two conflicting performances indexes.