A Proposal of Stochastic Quantitative Resilience Index Based on SLAs for Communication Lines

A wide-area distributed application is affected by network failure due to natural disasters because the servers on which the application operates are distributed geographically in a wide area. Failure Injection Testing (FIT) is a method for verifying fault tolerance of widely distributed applications. In this paper, by limiting network failures only to the connection line, whole FIT scenarios are generated, and exhaustive evaluation of fault tolerance is performed. The authors propose a method to evaluate the resilience quantitatively by specifications of the distributed application and their service level agreements (SLAs) that consist of the distributed network. The authors compare the resilience of several distributed applications in several topologies and show returns of investigation (ROI) in each application and topology.