On stability regions in opportunistic scheduled-packet access networks

Opportunistic scheduling has widely been used in wireless networks to send packet data over fading channels. In the literature, scheduling performance was usually analyzed by assuming a fixed number of active users and infinite packet buffers. However, the assumption may not be valid due to traffic dynamics in opportunistic packet transmissions. In this paper, we analyze opportunistic scheduling systems with varying number of users and burst packet arrivals. In particular, we define a stability region as the union of a set of convex polyhedral regions. We investigate the tradeoffs between throughput and fairness in terms of stability region. First, the maximal aggregated rate is derived under an optimal scheduling policy, which only depends on the quality-of-service (QoS) requirements without channel-state information (CSI). Second, the perfect fairness limit is defined such that the aggregated rate below the limit can equally be shared among users. Third, the fairness is optimized under a given overall throughput. As a metric to reflect the QoS requirements, we also derive a closed-form average delay for a symmetric system. Theoretical analysis and numerical results demonstrate the advantages of opportunistic scheduling.