A Fairness-driven Active Queue Management Algorithm with Hash Table and Circular Buffer
Siqiao Hu, Jinsheng Sun, Qin Xu, Jianshou Kong · 2020
The output from research on fairness-driven Active Queue Management (AQM) in the past two decades is undoubtedly plentiful, which spans from stateless to stateful and from mild regulation to strict discipline. Among the numerous current proposals, the stateless algorithm distinguishes itself for its simplicity and being less demanding for processing and storage resources. One notable example of stateless AQMs is CHOKe along with its many variants, all of which depend on comparing incoming packets from candidates picked from the buffer queue. However, these schemes cannot work properly if the packet buffer is too small. In that case the queue will quickly run out of packets from aggressive flows, and new packets from these flows will no longer punished. A fairness-oriented Active Queue Management algorithm named HTCB is proposed in this paper. The proposed algorithm compares incoming packets from candidates picked from a circular buffer instead to identify potentially aggressive flows from unknown ones. Once a flow is identified its information will be registered in a hash table, and its packets will be punished directly if it is deemed as aggressive or exempted if it is regarded as compliant. Simulation results are presented to confirm the effectiveness and efficiency of HTCB in punishing aggressive flows by comparison.