FairSplit - An efficient near-optimal bandwidth splitting strategy for OFDMA in IEEE 802.11ax

IEEE 802.11ax introduces OFDMA where multiple users transmit concurrently. This is particularly useful in dense scenarios where the density of WiFi connected devices is very high in settings such as stadiums, airports, classrooms, etc. The effectiveness of such a system relies heavily upon an efficient Scheduling and Resource Allocation (SRA) algorithm. One approach towards designing such an algorithm is to model this problem as a constrained optimization problem that maximizes a linear utility function. Such a framework attempts to adapt several standard SRA rules, like maximizing the Sum Rate, Proportional Fair, to the OFDMA settings. State-of-the-art (SOTA) SRA heuristics lack theoretical guarantees. In this paper, we show a reduction of the SRA problem in OFDMA to a budgeted variant of the maximum weight matching problem on bipartite graphs. We observe that this problem is NP-hard. Hence, we design an efficient approximation algorithm, FAIRSPLIT, with provable guarantees. We demonstrate the effectiveness of our algorithm through a variety of simulations. We see that FAIRSPLIT outperforms SOTA by 11-48% in terms of aggregate throughput while utilizing the bandwidth more efficiently.