Scalable WiFi Media Delivery through Adaptive Broadcasts.

Current WiFi Access Points (APs) choose transmission parameters when emitting wireless packets based solely on channel conditions. In this work we explore the benefits of deciding packet transmission parameters in a content-dependent manner. We demonstrate the benefits specifically for media delivery applications in WiFi environments by designing, implementing and evaluating a system, called Medusa. In order to keep the APs relatively simple, we implement the Medusa functions in a media-aware proxy. More specifically, when forwarding our media traffic, Medusa requires that APs simply use the WiFi broadcast feature, and that they refrain from making decisions on which wireless packets to retransmit, or what PHY rates such packets should be transmitted at. Instead we combine these typical link layer functions with a few other content-specific choices, in the proxy. Through detailed experiments across diverse mobility and interference conditions we demonstrate the advantages of this scheme for both unicast and multicast media delivery applications. The advantages are particularly substantial in multicast scenarios, where Medusa was able to deliver a 20 Mbps HD video stream simultaneously to 25 clients, using a single 802.11 AP, with good to excellent PSNR.