PNOFA: Practical, Near-Optimal Frame Aggregation for Modern 802.11 Networks

MAC-layer frame aggregation has significantly improved the efficiency of IEEE 802.11n/ac networks by placing multiple MAC-layer data units in a large PHY-layer frame. In this paper, we focus on finding the optimal length of an Aggregated MAC Protocol Data Unit (A-MPDU) in order to maximize throughput. This problem has proved to be extremely challenging because of the chain of dependencies between consecutive A-MPDUs due to software retransmissions and because error rates can be higher in the later part of the A-MPDU. In this paper we develop a model of A-MPDU frame aggregation and use it to design a statistically optimal algorithm. We then develop a standard compliant, Practical, Near-Optimal Frame Aggregation algorithm (PNOFA). Our trace-based evaluation shows that across a variety of devices and scenarios PNOFA outperforms existing state-of-the-art algorithms and obtains throughputs that are within 97% of those obtained using the statistically optimal algorithm. Furthermore, we implement PNOFA on an 802.11ac Google Wifi access point. We find that when compared with the proprietary frame aggregation algorithm in the Qualcomm IPQ 4019 chipset's firmware, PNOFA increases average throughput by 17% in the scenarios tested.