Improving quality of service of TCP flows in strictly prioritized network

We propose a new congestion control scheme for parallel TCP flows with different QoS requirements in strictly prioritized networks. When a node opens multiple connections that are going to the same destination through strict-priority scheduler enabled networks, we can assume that there is a significant correlation of congestion events between the flows of different priorities. In this paper we leverage this correlation to improve QoS levels provided to flows. We propose an indicator-based parallel TCP and describe its implementation, TCP-I. The usefulness of our scheme and the benefit of congestion information sharing is shown when prioritized service is working in the network, and each TCP is assigned a priority depending on its QoS requirements. We show that the mixed QoS requirements of individual flows can be better served in strictly prioritized environment when TCP flows share other flow's congestion information. Further, we analyze the buffer size requirement at a strict-priority scheduler for the case of no admission control.