Usage of 802.11n in practice: A measurement study

Abstract

IEEE 802.11n offers several throughput enhancing features over its predecessor 802.11a/g. The two main features at the PHY layer are: MIMO and channel bonding, while the main throughput enhancing feature at the MAC/link layer is Frame Aggregation. While in theory, as well as in controlled experimental conditions, these features achieve throughput enhancements, the extent to which they are useful for in-the-wild deployments has not been studied thus far. This paper presents measurements from three sets of traces: one from a research conference, one from a busy airport, and the third from a dense classroom setting with extensive WiFi usage for classroom activities. Our findings are as follows: (a) the high data rates of 802.11n are not used substantially, although the presence of moderate rates is significant, (b) the use of the channel bonding feature is minimal in dense deployments, and (c) the percentage of bytes undergoing frame aggregation is considerable but the levels of aggregation are not very high. We also undertake controlled experiments which shed light on non-wireless, system bottlenecks. Specifically, we find that many clients are not equipped to handle high levels of frame aggregation. Worse, this interacts badly with the rate adaptation algorithm, significantly lowering the data rates being used. These issues need careful addressing before 802.11n features are used effectively.