ActiveKeeper: An Accurate and Efficient Algorithm for Finding Top-k Elephant Flows

Abstract

Finding top- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> elephant flows in high-speed network is a crucial task for network traffic measurement. Since modern routers do not hold enough on-chip memory to precisely track all flow sizes, identifying elephant flows becomes more challenging than ever. Existing solutions mostly use compact data structures to fit in limited on-chip memory. However, they have to use large-size counters, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , 20-bit counter, to prevent overflows when recording flow sizes. This wastes significant on-chip memory since flow size distribution is highly skewed. Most counters only require 8 bits to track mouse flows rather than 20 bits. In this letter, we propose ActiveKeeper, an efficient top- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> finding solution that uses two-mode active counter to record flow sizes, reducing memory usage while ensuring high accuracy. We also introduce dynamically exponential decay to increase the precision in expelling the small flows while holding the large ones. Experimental results based on real-world Internet traces show that our method outperforms the state-of-the-art by achieving high accuracy and higher precision with small on-chip memory usage.