A dependency-aware QoS system for mobile satellite communication

Deploying reservation-based Quality-of-Service systems in variable environments like mobile satellite communication requires networks to actively adjust the status of reservations such that they fit the currently available capacity. Communication in such a system normally consists of multiple interdependent reservations. Even something as simple as a single file download via TCP already requires two paths: a broad downlink (data) and a narrower uplink (acknowledgments). Neither of these works without the other. A smart QoS system takes these dependencies into account when allocating resources. In this paper we show that if reservations are handled independently from each other, capacity may be allocated to paths applications cannot use due to the inherent structure of the protocols used. We present a system to model relations, enabling the network to intelligently adapt a set of reservations to a change in transmission capacity. We analyze two communication scenarios and compare how relation-aware and non-aware fitting algorithms react on a decrease in the link capacity. We show that our relation-aware design is able to improve on wasted resources and decrease broken sessions.