The impact of the buffer unit on the performance in 802.11 wireless LANs

A lot of work has been devoted to the problem of dimensioning buffers at routers/switches or access points in the Internet. In this paper, we address the issue of buffer unit (granularity), and discuss its impact on the performance of scheduling policies over 802.11 Wireless LANs. We investigate the bottleneck link capacity sharing at the buffer of an access point under two key workload scenarios: long lived flows and a mix of short lived and long lived flows featuring an heavy-tail distribution. We represent the buffer granularity in byte or in packet, and highlight their impact by exploring the behaviors experienced by transfers with opposite directions with a variety of schedulers: SCFQ, FIFO, LAS, LASACK, Run2C etc.