Memento: Making Sliding Windows Efficient for Heavy Hitters

Cloud operators require timely identification of Heavy Hitters (HH) and Hierarchical (HHH) for applications such as load balancing, traffic engineering, attack mitigation. However, existing techniques are slow in detecting new heavy hitters. In this paper, we present the case identifying hitters through sliding windows . Sliding windows quicker more accurate to detect than current interval-based methods, but date had no practical algorithms. Accordingly, introduce, design, analyze xmlns:xlink="http://www.w3.org/1999/xlink">Memento family sliding window algorithms HH HHH problems single-device network-wide settings. We use extensive evaluations show that our solutions orders magnitude faster comparable speed state-of-the-art non-windowed sampling based technique. Furthermore, exemplify detection capabilities on a realistic testbed. To end, implemented Memento an open-source extension popular HAProxy cloud load-balancer. evaluations, using HTTP flood by 50 subnets, approach detected subnets reduced number undetected requests up $37\times $ compared alternatives.