Comparing the Performance of Inter-Sector/Intra-Sector Scheduling and ARQ for Multimedia Traffic in Wireless Access Networks This work has been supported by the National Capital Institute of Telecommunications (NCIT), Ottawa, Canada. Mohamed H. Ahmed Faculty of Engineering and Applied Science, Electrical and Computer Engineering, St. John’s, Canada mhahmed@engr.mun.ca Halim Yanikomeroglu and Samy Mahmoud Broadband Communication & Wireless Systems (BCWS) Centre Dept. of Systems & Computer Engineering Carleton University, Ottawa, Canada {halim, mahmoud}@sce.carleton.ca Abstract-Multimedia transmission in wireless networks requires high quality of service in terms of packet error rate (PER), throughput and packet delay. Inter- Sector/Intra-Sector Scheduling (ISISS) has been proposed [1] to use basestations coordination in packet scheduling to reduce PER by avoiding (minimizing) concurrent transmission of potential cochannel interferers. Since ARQ schemes can also be used for PER reduction, it is imperative to compare the performance of ISISS with that of ARQ. Results show that ARQ is more effective in reducing PER at the expense of the incurred packet delay due to packet retransmission particularly at medium to high loading. Keywords-ARQ, packet scheduling, multimedia wireless communications. I. INTRODUCTION Inter-Sector/Intra-Sector Scheduling (ISISS) concept has been introduced in [1] as a dynamic time slot allocation that enhances the packet-level system performance by minimizing the cochannel interference. In ISISS, basestations (BSs) exchange information of the available traffic. Then, each BS schedules its local traffic based on this information to avoid concurrent transmission by potential dominant interferers. Therefore, ISISS schedules the packet transmission in each sector taking into account the traffic information in the sectors of potential dominant interferers. For instance, as shown in Fig. 1, the potential dominant interferers for users in sector 1(shaded sector)/cell 4, are the signals transmitted for users in sector 1/cell 1 and sector 1/cell 7. It is noteworthy that users in sector 1/cell 7 cause this interference in the downlink because of the assumed wraparound grid. Therefore, BS 4 sends the traffic information of users in sector 1/cell 4 to BSs 1 and 7. Meanwhile, BS 4 receives the traffic information of users in sector 1/cells 1 and 7. This information includes the arrival time and service type of each packet waiting in the transmission queue. BSs 1, 4, and 7 use this information to avoid (or minimize) concurrent transmission of any users in these three sectors (sector 1/cells 4, 1, and 7). We call these sectors that have potential dominant interferers “interference group”. By eliminating the concurrent transmission within each interference group, the interference level can be drastically reduced. Obviously, there are other users that still can cause interference to users in sector 1/cell 4 including users in sector 1/cells 2, 3, 5, 6, and 9. However, their interference level is much less than that of sector 1/cel1 1 and 7 because of the large path loss and the antenna directivity at both BSs and SSs. The number of occupied slots in each sector as well as the decision of which time slots are allocated to which users in each sector depend on the employed intra-sector and inter-sector scheduling schemes. The intra-sector scheduling scheme schedules the packet transmission of all users inside a sector while the inter-sector scheme schedules the traffic transmission of different sectors within the interference group as shown in Fig. 2. First Come First Serve (FCFS), Weighted Round-Robin (WRR), Weighted Fair Queuing (WFQ), or any other schemes can be used at any of the two levels. It has been shown in [1] that ISISS can reduce the packet error rate (PER) by one order of magnitude but at the expense of the packet delay. Automatic Repeat Request (ARQ) schemes can also be used to decrease PER. However, ARQ is expected to cause significant packet delay and throughput degradation. Hence, it is essential to compare the performance of both schemes