Designing Network Monitoring Schemes Based on Segment Routing

Abstract

Network monitoring is important to guarantee the reliable end-to-end communications. Segment routing can be used to reduce network monitoring cost. The associated minimum cycle cover problem is to find a set of cycles such that (a) all cycles pass through a given monitor/source node, (b) each link in the network belongs to at least one cycle, (c) each cycle has a segment list no bigger than a given threshold K, and (d) the cycle cover length is minimized. To the best of our knowledge, SCMon is the first network monitoring scheme based on segment routing. But instead of directly minimizing the cycle cover length, SCMon focuses on reducing the number of cycles required. In this paper, four new algorithms, denoted by A, B, C and D, are proposed to solve the minimum cycle cover problem. Among them, algorithm A is a simple extension of SCMon by focusing on reducing the cycle cover length. Algorithm B adopts a new notion of symmetric cycles. Algorithm C enhances B by using a new path metric for constructing cycles. Algorithm D enhances C with a parallel cycle construction process. We show that all four algorithms outperform SCMon and when K is large, algorithms B, C and D can always find optimal solutions.