Learning Based Adaptive Fair QoS in IEEE 802.11ac Access Networks

IEEE 802.11ac for high throughput wireless access supports the use of wide channel bandwidth for improving the capacity of data transmission. To properly balance the needs of individual wireless stations (STAs) and the available network capacity, the standard proposes dynamic bandwidth channel access (DBCA) where the wide channel is divided into primary and secondary channel components; and a station needs to contend for the primary channel while it can use the secondary channel without any contention, when it is sensed free. This leads to the unfairness in channel access. Further, the network gets severely affected when STAs demand for specific Quality of Service (QoS) for various applications belonging to different access categories (ACs) running over it. Different ACs with different QoS types require separate treatment in terms of channel bandwidth allocation, which the standard IEEE 802.11ac DBCA fails to provide. In this paper, we develop a packet scheduling mechanism, High Throughput Multi-Level Scheduler (HT-MLS), to support dynamic bandwidth allocation over QoS associated traffic classes, while ensuring the fairness among the contending STAs. HT-MLS utilizes an adaptive learning over the channel sensing environment; and a testbed implementation (with 30 wireless stations) and thorough evaluation of HT-MLS indicate that it can improve QoS performance significantly, while ensuring the channel access fairness for the STAs.