Detection of multipath routing with passive delay measurements: Hypothesis testing approaches. Issue 8 (22nd October 2019)

• Record Type: Journal Article
• Authors: Pan, Shengli; Wang, Yan; Bai, Ziming; Xiong, Muzhou; Gu, Lin
• Is Part Of: Transactions on emerging telecommunications technologies. Volume 33:Issue 8(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2019-10-22
• Subjects: Telecommunication -- Periodicals; 384.05
• DOI: 10.1002/ett.3758
• Languages: English
• ISSNs: 2161-5748

Abstract

Abstract: Multipath routing, which is a common approach to achieve load balancing, will enable multiple active paths between two end‐hosts. However, it will cause out‐of‐order delivery of packets, which will not only severely debilitate the performances of TCP (Transmission Control Protocol) but also will make many single‐path–based applications work improperly. In this paper, we propose to utilize passive delay measurements between end‐hosts, to detect multipath routing with two different hypothesis testing approaches, ie, t‐Test and variance test. The motivation is that the obtained distributions of passive delays between end‐to‐end communication flows can be very similar to each other when they are routed by the identical end‐to‐end path, whereas distinct differences are observed among the ones of different end‐to‐end paths. Simulations based on NS2 validate the efficiency of both t‐Test and variance test and demonstrate the robustness of t‐Test against asynchronous measurements between different flows. Abstract : We propose to utilize passive delay measurements between endhosts, to detect multipath routing with two different hypothesis testing approaches, ie, t‐Test and variance test. The motivation is that the obtained distributions of passive delays between end‐to‐end communication flows can be very similar to each other when they are routed by the identical end‐to‐end path, whereas distinct differences are observed among the ones of different end‐to‐end paths. Simulations based on NS2 validate the efficiency of both t‐Test and variance test and demonstrate the robustness of t‐Test against asynchronous measurements between different flows.