Improved Utility-based Congestion Control for Low-Delay Communication
This paper proposes a novel congestion control algorithm for low-delay communication over best effort packet switched networks. Due to the presence of buffers in the internal network nodes, each congestion leads to buffer queueing and thus to an increasing delivery delay. It is therefore essential to properly control congestions in delay-sensitive applications. Delay-based congestion algorithms could offer a viable solution since they tend to minimize the queueing delay. Unfortunately they do not cohabit well with other types of congestion algorithms, such as loss-based algorithms, that are not regulated by delay constraints. Our target is to propose a congestion control algorithm able to both maintain a low queueing delay when the network conditions allows for it and to avoid starvation when competing against flows controlled by other types of policies. Our Low-Delay Congestion Control algorithm exactly achieves this double objective by using a non-linear mapping between the experienced delay and the penalty value used in rate update equation in the controller, and by combining delay and loss feedback information in a single term based on packet interarrival measurements. We provide a stability analysis of our new algorithm and show its performance in simulation results that are carried out in the NS3 framework. They show that our algorithm compares favorably to other congestion control algorithms that share similar objectives. In particular, the simulation results show good fairness properties of our controller in different scenarios, with relatively low self inflicted delay and good ability to work also in lossy environments.
Record created on 2015-01-31, modified on 2016-08-09