Client devices must cooperate to share limited resources in wireless local area networks (WLANs). Standards exist for access points (APs), to use a more holistic view of the RF environment, to share information with clients (802.11v/k), allowing clients to make better performance-impacting decisions, e.g. choosing to roam between APs. Standards intended to lower the cost of these expensive control-plane operations (802.11r) also exist. In this paper, we use large scale measurements to show the limited adoption of existing client-based techniques. We propose ClientMarshal for enterprise WLANs, which provides network infrastructure more control over these decisions without modification to client device software. ClientMarshal divides the enterprise network into logical groups called cells, each consisting of a small number of neighboring APs operating on different channels. All APs belonging to the same cell broadcast the same wireless identity (BSSID) to client devices, and appear as a single virtual AP. The key enabler behind ClientMarshal is FreeSteer (FST), a novel protocol we develop that allows an AP to dynamically steer an associated client to another AP in the same cell based on various network measurements in order to improve overall network performance. This design also improves the roaming experience by bypassing traditional control-plane operations, avoiding significant overhead as well as client implementation defects. Using systematic evaluation and implementation on commodity APs, we show the efficacy and control of FST protocol in client steering. Compared to state-of-the-art approaches, FST can steer clients almost 2.5X faster, with 52% higher success rate, and reduce packet loss by more than 68%.

Agent-level negotiations between an application and a cloud service provider (SP) need to consider two complexities: i) inaccuracy of computational models in capturing the behavior of a cloud-based system made up of diverse components and resources; and ii) disparity in the goals and priorities of an application and the SP in exercising a cloud-based system. The paper suggests the use of trial actions by an agent to learn about the system behavior and the opponent agent goals, as part of an effective negotiation strategy. The trial actions allow searching, alternations, and improvements on the system QoS space, towards a larger goal of optimal QoS control.