Eighth ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD 2007)
LQRD: An Improved ECN Algorithm
Haier International Training Center, Qingdao, China
July 30-August 01
ISBN: 0-7695-2909-7
In this paper the performance of the gateway using ECN (Explicit Congestion Notification) is analyzed firstly. Study shows that the recommendation in the protocol that gateway drops packets when the average queue size exceeds the upper threshold brings too low throughput. However, if it is be substituted for marking packets, which increases average delay and induces parts of fail connections for heavy loads, the throughput will improve greatly. To make sure high throughput, low delay and robust connectivity for long queue length, this paper presents an improved algorithm named LQRD (Long Queue Random Drop). The algorithm drops packets with a certain probability when the average queue size exceeds the upper threshold to limit the increase of queue length. The drop probability is computed according to buffer size, link capacity, average queue length and average queue delay to reflect load changing. Then we analyze the solved model of the LQRD algorithm by using differential equation. And at last, some simulation experiments show the algorithm has good performance.
Citation:
Jun Lai, Wu Ye, Sui-li Feng, "LQRD: An Improved ECN Algorithm," snpd, vol. 1, pp.239-244, Eighth ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD 2007), 2007