Analysis of Temporal Latency Variation on Network Coordinate System for Host Selection in Large-Scale Distributed Network

A network coordinate system (NCS) which predicts latencies between end-computers at low cost can achieve efficient resource selections on a large-scale distributed system such as a cloud computing. The NCS first decides coordinates of each computer on the pre-determined geometric space, and then estimates end-to-end latency by calculating a distance between the coordinates. Therefore, temporal variation of the latencies on a dynamic network environment affects the prediction accuracy of the NCS. In this paper, we build a mathematical model which can accurately represent the latency variation by applying time series analysis with ARIMA model to the dataset obtained from global network measurement services. And then, the impact of the realistic latency variation on the performance of the NCS is evaluated through computer simulations. From the evaluation results, we have clarified that the time variation of the latencies rapidly degrades the prediction accuracy of the existing NCS algorithm.