Performance evaluation of proportional fair scheduling algorithm with measured channels

Abstract

Motivated by the promising performance results of the proportional fair (PF) packet scheduling algorithm, often quoted in connection with WCDMA High Speed Downlink Packet Access (HSDPA), we have performed an evaluation of the PF gain in comparison to the simpler round robin (RR) scheduler when subjected to measured channel traces. Specifically, we applied measured signal fading recorded from GSM cell phone users making calls on an indoor wireless office system. Different from reference channel models, these measured channels have much more irregular fading between users, which as we show by simulation leads to vanishing system throughput gain of PF over RR; the maximum throughput (Max TP) scheduler on the other hand, is much less influenced by the measured channels, but also less fair in giving equal throughput to the users, as is the RR scheduler in comparison to PF.