, Franklin Liu, An Wang, Hang Liu
Identifying heavy ﬂows, i.e., ﬂows with high packet rate (or count), is vital for many network applications such as anomaly detection, network operation, quality of service, etc. The task of real-time heavy ﬂow detection in data plane is challenging due to high switching speed (100 Gbps), a large number of concurrent ﬂows (millions of concurrent ﬂows), and small memory footprint requirement. In this paper, we dissect the key factors that affect the existing detection schemes accuracy, and propose AccuPipe, a new detection scheme with intelligent ﬂow entry replacement strategies. The experiment results show that the new scheme is able to efﬁciently utilize all ﬂow entries in the detection pipeline, and detects more than 850 heavy ﬂows (out of top 1,000) using a small amount of memory (1,000 ﬂow entries, roughly equivalently to 18KB memory) in a programmable switch, a 70% improvement over the existing scheme. In addition, we investigate the performance of the proposed scheme with varying ﬂow entry replacement strategies, and report their pros and cons.
April 20-24, 2020
IEEE/IFIP Network Operations and Management
Heavy Flow Detection, Network Measurement, Programmable Switch, Software Defined Networking