SRLA: A real time sliding time window super point cardinality estimation algorithm for high speed network based on GPU

TL;DR

This paper introduces SRLA, a GPU-based algorithm for real-time super point cardinality estimation in high-speed networks using sliding time windows, achieving high accuracy and low latency.

Contribution

SRLA is the first algorithm to efficiently estimate super point cardinality under sliding time windows with incremental updates and GPU parallelization.

Findings

Estimates super point cardinality within 100 ms on a GTX650 GPU.

Outperforms existing algorithms with over 20 times faster estimation.

Successfully applied to real-world traffic with 40 GB/s bandwidth.

Abstract

Super point is a special host in network which communicates with lots of other hosts in a certain time period. The number of hosts contacting with a super point is called as its cardinality. Cardinality estimating plays important roles in network management and security. All of existing works focus on how to estimate super point's cardinality under discrete time window. But discrete time window causes great delay and the accuracy of estimating result is subject to the starting of the window. sliding time window, moving forwarding a small slice every time, offers a more accuracy and timely scale to monitor super point's cardinality. On the other hand, super point's cardinality estimating under sliding time window is more difficult because it requires an algorithm to record the cardinality incrementally and report them immediately at the end of the sliding duration. This paper firstly solves this problem by devising a sliding time window available algorithm SRLA. SRLA records hosts cardinality by a novel structure which could be updated incrementally. In order to reduce the cardinality estimating time at the end of every sliding time window, SRLA generates a super point candidate list while scanning packets and calculates the cardinality of hosts in the candidate list only. It also has the ability to run parallel to deal with high speed network in line speed. This paper gives the way to deploy SRLA on a common GPU. Experiments on real world traffics which have 40 GB/s bandwidth show that SRLA successfully estimates super point's cardinality within 100 milliseconds under sliding time window when running on a low cost Nvidia GPU, GTX650 with 1 GB memory. The estimating time of SRLA is much smaller than that of other algorithms which consumes more than 2000 milliseconds under discrete time window.