Experimental investigation of the impact on TCP traffic by hidden video-tagged UDP traffic

Abstract

This paper investigates the impact of video-tagged UDP traffic on best-effort TCP traffic in an experimental setup with a novel observation approach. Our setup consists of three 802.11ac nodes in a hidden node topology, where a 20Mbps UDP and a best effort TCP traffic flow from the edge (hidden) nodes to the center node. We observe that, even though the RTS/CTS mechanism is enabled, when UDP and TCP traffic flow simultaneously, video-tagged UDP traffic causes the besteffort TCP traffic throughput to drop by 50%. An analysis of the captured and synchronized packet data reveals for the first time that, the following event is a surprising major contributor to the throughput degradation. Since the TCP and UDP sources are hidden from each other, an RTS packet of the UDP source may collide with the CTS reply to the TCP source. This in turn may result in many collisions between the subsequent RTS packets of the UDP source and the data packet of the TCP source. This paper experimentally shows that this event is particularly commonplace for video-tagged traffic, having smaller back-off duration, severely impacting the cross-traffic.