Hybrid Energy Efficient Dynamic Bandwidth Allocation in Long Reach TDM-PONs

• Main Authors: Shi-JieLin, 林士傑
• Other Authors: Chuan-Ching Sue
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/05623732499987252225

碩士 === 國立成功大學 === 資訊工程學系 === 102 === Abstract –Energy saving is an essential issue in passive optical networks (PONs) due to the increased concerns for global warming. Among the various energy saving techniques in PONs, putting the idle optical network unit (ONU) into sleep mode to conserve power is considered as a promising approach. Therefore, many sleep mode scheduling mechanisms have been proposed based on the ONU sleep mode. The previously proposed mechanisms can be broadly classified as either sleep for more than one cycle mechanisms or sleep within one cycle mechanisms. The two categories mainly differ in their scheduling policies. In this thesis, we mainly focus on improving sleep within one cycle mechanisms. In sleep within one cycle mechanisms, the OLT would poll (i.e., transmit a GATE message) each ONU in every cycle where the GATE message contains information to the ONU, including the transmission window and the next wake-up time. The next wake-up time is determined by the next polling time. In fact, the OLT may not be able to determine the next polling time at the current polling time. In previous studies, Sleep Mode Awareness (SMA) uses a prediction method to find the time epoch. However, prediction error may induce ONU early wake-up and result in an extra power consumption which decreases the power-saving efficiency of the system. Furthermore, the extra power consumption will increase when the system load and the round trip time (RTT) increase. We propose a sleep scheduling mechanisms named Single GATE Sleep Scheduling (SGS) which can avoid the ONU early wake-up problem by splitting one sleep period into two sleep periods, but it will cause a fixed extra power consumption due to extra mode switching overhead. To minimize the extra power consumption of the system, we develop an effective scheduling method of combining our proposed method with SMA, called Hybrid-SGS-SMA mechanism. By comparing the fixed extra power consumption with extra power consumption caused by prediction error, Hybrid-SGS-SMA can dynamically select sleep scheduling mechanism according to the current environment. The simulation results show that Hybrid-SGS-SMA can directly switch between different sleep scheduling mechanisms without reconfiguration overhead and thus minimizing the power consumption of the system.