A virtual loss-load congestion control strategy for high speed networks

Narayanan Prithviraj, Carey L. Williamson
1996 Computer communication review  
Virtual Loss-Load model is a hybrid congestion control strategy which combines the leaky bucket traffic shaper with lossload model to get higher network utilization and reduce network congestion. Two performance parameters used in the simulation are the file transfer time and the packet loss probability. The transmitted file has finite length instead of infinite. The parameters of the virtual loss-load model can be tuned to trade off the packet loss and the file transfer time. The simulation
more » ... ults show that the virtual loss-load scheme is comparable to several other reactive congestion control schemes. 2003: KUNNIYUR, S., AND R. SRIKANT End-to-end congestion control schemes: utility functions, random losses and ECN marks. IEEE/ACM Transactions on Networking 11, 5, (Oct 2003), 689-702. Abstract: The authors provide a simple framework based on deterministic fluid models to study the effect of utility function, random losses, and explicit congestion notification on the design of the end-to-end congestion control scheme. First, the authors show that there exists an additive-increase-multiplicative-decrease scheme that uses end-to-end measurable losses to convergence to the socially optimal solution. Second, the authors incorporate the round-trip delay and get more general window control schemes. Third, the ECN is used as congestion indictor instead of packet losses. Decentralized marking levels can be chosen to achieve near zero-free service. KESSELMAN, A., AND Y. MANSOUR Adaptive AIMD congestion control. Abstract: The authors provide an adaptive increase/multiplicative decrease (AIMD) congestion control scheme to achieve the high bandwidth utility, fairness to the connections, and responsiveness in the changing environment. The increase item and decrease factor are dynamically updated based on the feedback. The authors show that both single connection and multiple connections can converge to a steady state approximating the optimal utilization from any arbitrary state, and multiple connections can converge to the fair share. But the algorithm is not TCP-friendly. SHAKKOTTAI, S., R. SRIKANT, AND SEAN P. MEYN Bounds on the throughput of congestion controllers in the presence of feedback delay. IEEE/ACM Transactions on Networking 11, 6, (Dec 2003), 972-981. Abstract: The authors study a single node with multiple flows based on a deterministic fluid model where the multiple flows have different round-trip delays. The results show the total rate at the router is bounded and the total rate grows at most linearly with the increase of the system size. Even though the linearized model is not stable, the total rate at the router is upper bounded. In the study, we assume that the queueing delay at the router is ignorable compared to the propagation delays.
doi:10.1145/231699.231704 fatcat:sjyddm4bbzgfxc5chlus7mf4le