Energy-Efficient Dynamic Bandwidth Allocation for Long-Reach Passive Optical Networks Dung Pham Van 1,2 , M. Pubuduni Imali Dias 1 , Koteswararao Kondepu 2 , Luca Valcarenghi 2 , Piero Castoldi 2 , and Elaine Wong 1 1 Department of Electrical & Electronic Engineering, The University of Melbourne, VIC 3010, Australia. 2 Scuola Superiore Sant’Anna, Pisa, Italy Email: dung.pham@unimelb.edu.au Abstract For the first time, an energy-efficient dynamic bandwidth allocation scheme for long reach PONs is proposed. Results show that the proposed scheme significantly saves ONU energy whilst incurring acceptable frame queuing delays. Introduction Long-reach passive optical networks (LR-PONs) are one of the major trends in the evolution of optical access- metro networks. In contrast to traditional PONs, LR- PONs suffer from long propagation delays due to extended network reach between the OLT and ONUs. Its long round trip time (RTT) greatly impacts the overall frame delay including the time between the frame arrival and the REPORT message sent for it (polling delay) and that between the REPORT message and its corresponding GATE message (granting delay) [1]. From the energy efficiency’ perspective, however, such long delays can be exploited to schedule ONU power saving modes, i.e., sleep or/and doze mode [2] to improve energy-savings. To the best of the authors’ knowledge, no research has been reported to date on incorporating power saving mechanisms with dynamic bandwidth allocation (DBA) algorithms in LR-PONs. This paper proposes an energy- efficient DBA (EDBA) scheme that aims at maximizing ONU energy-savings in long reach TDM-PONs, i.e., LR-PONs with a single shared upstream (US) wavelength and a single shared downstream (DS) wavelength. The proposed EDBA scheme is based on the sleep aware dynamic bandwidth allocation (SDBA) scheme presented in [3] with new features introduced to improve ONU energy-savings in LR-TDM-PONs. Energy-efficient dynamic bandwidth allocation Fig. 1 illustrates the operation of the proposed EDBA scheme. For illustration purposes, only 2 ONUs are considered. Given a cycle time T c , all ONUs are assigned the same timeslot T slot to ensure fairness amongst them. Within an assigned timeslot, data transmission including both DS and US transmission directions takes place followed by control message transmission. T c is determined so that the transmission slot for both data and control messages Tx_len is upper bounded by T slot . Depending on the DS and US traffic load, Tx_len is minimized. Outside the Tx_len slot, the whole ONU transceiver is switched off, i.e., sleep mode, for saving energy. From the time a REPORT is sent until when its corresponding GATE is received (granting delay time), only ONU transmitter is switched off, i.e., doze mode, for further energy reduction. The operation of EDBA consists of an offline initialization phase and an online phase. As shown in Fig. 1, during the offline phase, the OLT measures and informs all ONUs their RTTs, whilst each ONU reports its US buffer backlog. Once all the initial REPORTs are received, the OLT performs off-line scheduling, for the first normal polling cycle, i.e., generating and sending GATEs to all ONUs in sequence. The EDBA operation during the online phase is described as follows: Step 1: When an ONU receives a GATE, it first reassigns its local clock onu_clk to the OLT clock olt_clk for synchronization purposes [3]. Then, it extracts the next transmission slot start time Tx_start and duration Tx_len. US traffic is buffered until Tx_start. The ONU sleeps only if the idle time T idle = Tx_start - onu_clk is larger than the sleep overhead time Toh s [4]. Step 2: When onu_clk = Tx_start - Toh s , the ONU takes Toh s time to wake up and prepare its transceiver for data and message transmission. Step 3: When the local clock reaches Tx_start, both OLT and ONU transmit buffered DS and US data, respectively. The OLT transmits DS traffic until olt_clk = Tx_start + Tx_len - T msg , then it generates a new GATE message and sends to the ONU. Here, T msg is the time for processing a GATE message and a REPORT message in a timeslot [3]. Step 4: When either the ONU US data buffer is empty or onu_clk = Tx_start + Tx_len - RTT - T msg , the ONU sends a REPORT message containing updated US buffer backlog to the OLT. After that, the ONU switches off its transmitter whilst leaving its receiver active to receive DS traffic, i.e., in doze mode operation. Step 5: Once the ONU receives the new GATE signifying the end of the DS transmission, it extracts the new Tx_start and Tx_len, and then switches off also its receiver, i.e., transitioning from doze mode to sleep mode operation. The protocol operation is then repeated. The major difference between the EDBA and its predecessor SDBA [3] is the doze mode implementation during the REPORT-to-GATE time for further improvement in energy-savings. This is based on the fact that DS traffic is usually heavier than US traffic, and 978-1-922107-21-3  2014 Engineers Australia OECC / ACOFT 2014 ● 6 -10 July 2014 ● Melbourne, Australia 999