Lightweight Per-Flow Traffic Measurement Using Improved LRU List

Abstract

Per-flow traffic measurement is extremely important for various applications, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> , flow size distribution, flow cardinality and DDos detection. Previous works, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> , Elastic Sketch and FlowRadar, incur massive per-packet processing overhead and/or large memory consumption. In this paper, we present a novel least-recently-used (LRU) list based solution, called LRU-TX, for lightweight and accurate per-flow traffic measurement. Different from the classic LRU list with one exit at the tail, we maintain another exit at a certain position <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$r$</tex-math></inline-formula> (not the tail) in the list. Many terminated flows will be removed from the exit at position <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$r$</tex-math></inline-formula> as soon as possible with their counters less than a given threshold, which helps to decrease both the amortized processing overhead and the memory consumption. We also formally analyze the impact of different parameters ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> , the exit's position, the threshold) on the measurement performance. Experimental results and extensive simulation results show that LRU-TX can increase the throughput by 30% at least, and reduce the memory cost by 50%, compared with the state-of-the-art solutions under the same measurement error conditions.