Reliable Redundant Services Placement in Federated Micro-Clouds

Abstract

In a battlefield or disaster relief environment, network connectivity ability is unstable and task requirements may change frequently. Federated micro-clouds are promising for solving problem of service delivery in this dynamic scenario. However, due to the harsh environment and the dynamic feature of federated micro-clouds, there is a serious risk of service failure. Redundant configuration is an effective method to enhance service reliability. But when determining the placement strategy for an application request containing multiple services, few previous approaches can effectively handle such a highly complex task of redundant services placement under multiple constraints, including resource constraints, communication delay limits and location constraints. In this paper, we study how to reliably place redundant services requested by user on federated micro-clouds. We first design a two-level reliability evaluation model to quantitatively measure the reliability of each services placement from the two aspects of node reliability and Shared Risk Domain (SRD) reliability. Second, we propose a Reliable Redundant Services Placement (RRSP) approach to iteratively place each service by using Dynamic Graph Growth (DGG) algorithm. RRSP can determine the current optimal location according to the reliability of nodes and the number of service combinations to find a highly reliable placement strategy in a short time. The evaluation results show that our algorithm can provide more reliable services placement than three other representative algorithms.