Lee, SanghwanZhang, Zhi-LiNelakuditi, Srihari2020-09-022020-09-022004-02-20https://hdl.handle.net/11299/215602Multi-homing, namely, connecting to more than one Internet Service Provider (ISP) for global Internet reachability, is a common practice among many (especially large) customer (or stub) networks. Although the purpose of multi-homing is primarily for enhanced reliability, it has also increasingly been used for load balancing and other performance benefits. This paper is motivated by the following major question: in a multi-homed stub network, is there any significant benefit in carefully selecting one of the several available ISPs to optimise latency (as measured by round trip time, RTT) to various destinations? To answer this question, we carry out a measurement-based study to compare and analyze performance differences in using two different providers in a multi-homed stub network to reach a large number of randomly selected destinations. Our study reveals that there are often performance benefits in selecting the best provider to optimise network latency. Furthermore, for a large fraction of the network prefixes, the RTT differences between the two providers fall into a dominant range. This phenomenon can be attributed to the effect of the AS hierarchy on AS paths: the AS hierarchy often causes the AS paths via the different providers to merge at the core of the Internet, resulting in shared common segments to many network prefixes and ASes. Consequently there is strong correlation among RTTs to many destination networks. Our findings provide some useful insights as to how to perform intelligent provider selection using BGP in a multi-homed stub network.en-USReport