Acknowledge-Based Non-Congestion Estimation: An Indirect Queue Management Approach for Concurrent TCP and UDP-Like Flows

Abstract

This paper presents a new approach for indirect Active Queue Management (indirect AQM) technique called Acknowledge-based Non-Congestion Estimation (ANCE), which employs end-to-end queue management along a network instead to use Explicit Congestion Notification (ECN) bit or to drop packets in the queue. The ANCE performance was compared with Random Early Detection (RED), Control Delay (CoDel), Proportional Integral controller Enhanced (PIE), Explicit Non-Congestion Notification (ENCN), TCP-Jersey and E-DCTCP schemes in a daisychain and in a dumbbell cenario, with TCP flows and UDP-like Networked Control Systems (NCS) flow sharing the same network topology. On the other hand, this paper presents a method for modeling, simulation and verification of communication systems and NCS, using UPPAAL software tool, on which, all network components (channels, routers, transmitters, receivers, plants, and Controllers) were modeled using timed automata making easy a formal verification of the whole modeled system. Simulations and statistical verification show that despite using fewer resources (since ANCE does not need the ECN bit) ANCE presents a very close performance to ENCN overcoming Drop Tail, RED, CoDel, PIE and E-DCTCP in terms of Integral Time Absolute Error (ITAE) for NCS and fairness for TCP flows. ANCE also attains better performance than RED, PIE, TCP-Jersey and E-DCTCP in terms of throughput for TCP flows.