Active Queue Management in Internet and Wireless Networks X. Deng, S. Yi, G. Kesidis and C. R. Das The Pennsylvania State University ABSTRACT METHOD (CONT) In this research, we use a control theoretic approach to develop a generic Use Rate-based AQM for Uplink Traffic in Wireless LANs framework for analyzing various active queue management (AQM) • Ensure fairness for downlink/uplink traffic schemes as proportionally-integral-derivative (PID) controllers. Based on • Can work together with downlink queue-based AQM for Queue this PID model, we propose an adaptive control mechanism to improve management and QoS provisioning the system stability and performance under changing network conditions. AQM(Gc) p TCP (Gt) Queue (Gq) q A simulation study under a wide range of traffic conditions suggests that Downlink the proposed algorithms outperform the existing AQM schemes in achieving better system performance and stability. We then apply the -1 AQM concept to achieve fairness bandwidth allocation between downlink and uplink traffic in wireless LANs TCP (Gt) MOTIVATION p Uplink AQM(Gc) -1 Modeling TCP/AQM • A control-theoretic model for TCP and RED [Hollot’01]. • Extend this model to incorporate more end-system and AQM controller • Proportional-Integral-Derivative (PID) controller – A unified RESULTS framework for design and analysis of AQM schemes. KI • System stability is related to network and traffic condition – use • Putting everything together : GC K P s KDs Adaptive Control for a better stabilized and robust system AQM for Wireless Networks • More variable in bandwidth and delay • Uplink/Downlink asymmetry SYSTEM MODEL W q AQM Queue TCP N queue dynamic dynamic R FIFO • Stabilized queue size with varying traffic loads and link delays p • Reduced packet loss and response time time delay AQM R secs TCP-Reno AQM(Gc) p TCP (Gt) Queue (Gq) q Queue Controller Plant (Gp) -1 Linearized TCP/AQM System rate based •N: Number of connections •W: TCP congestion window •R: Round trip time •: Packet arrival rate •q: router queue size •p: packet drop/mark probability • Better Fairness between uplink/downlink traffic by using PI control METHOD over packet arrival rate (AVQ) for uplink traffic in Wireless LAN • A Unified Framework for Design and Analysis of AQM With AVQ No AQM • Interpreting and Calibrating AQM parameters as PID controlling components to the TCP plant module. • New Controlling Components in AQM • Derivative Control (D control): Use change of packet arrival rate over time as the congestion indicator, in addition to the instantaneous packet arrival rate (P control) and router queue length (I control). • Adaptive Control: The controlling parameters adaptively changes with the steady state packet marking probability p0 using update functions m and l, which increase monotonically (APID). q m( p) l ( p0 ) p Gc CONCLUSIONS Gc l ( p0 ) Gc • Virtual Queue Control: A New Combination of rate-based and • Various AQM Schemes can be classified as different types of PID queue-based control (QRVQ) 1 p0 C Controllers. p - q • Addition of derivative and adaptive control improves the stability of Gt Gq Cs the system under varying plant configurations. • A rate based AQM for uplink traffic in Wireless LANs enhances the -1 fairness and improves the QoS for downlink traffic.