Distributed measurement system for the end-to-end network performance using a dynamic job management

Abstract

The network monitoring and the analysis of the end-to-end performance is a critical task for Internet Service Providers (ISPs) that would like to estimate their subscribers' actual network performance to improve their service quality. Although many research studies for the network performance monitoring have been only focused on the Internet backbone, inter-AS connectivity with their monitoring systems, the study of a monitoring infrastructure for the distributed small clients like home appliances and PCs is quite few. In addition, ISPs would like to know the quality of their subscribers' network in real time. Since their subscribers are distributed and have client PCs with a very short period time of the network connection, we can't directly apply the traditional network measurement methods to the environment. This paper proposes a new approach for the end-to-end network measurement that is focused on the performance for ISP subscribers to the Internet hosts. Our contributions of the paper are as follows: 1. Design and development of a new approach for the end-to-end network performance monitoring with an dynamic program loading to distributed small clients like home appliances and PCs, and the analysis of the monitoring data is done at a central server. 2. The experiment to evaluate the proposed method on the operational ISP network based on the topology construction with the analysis of several links from clients.