A Probability-Based Fault Tolerance Strategy For Service-Based Systems

In the cloud environment, software vendors have built a large number of multi-tenant service-based systems (SBSs) by integrating distributed services to provide customers with software-as-a-service. However, when operating in the volatile cloud environment, the system components of an SBS may experience quality degradation caused by various runtime anomalies, e.g., virtual machine faults, software exceptions, etc. Handling runtime anomalies of multi-tenant SBSs timely and cost-effectively is critical to ensuring the quality of the SBS. A number of fault tolerance approaches have been proposed in recent years to tackle this problem. However, existing approaches have not properly considered the fact that the quality of a distributed system component often varies over time. In this paper, we propose PFT4MTS (Probability-based Fault Tolerance for Multi-Tenant SBSs), a novel approach for formulating cost-effective fault tolerance strategies for multi- tenant SBSs by implementing redundancy for critical system components. First, PFT4MTS evaluates the criticality of each system component by its multi-dimensional quality over time and multiple tenants sharing the system components with differentiated quality preferences. Then, it employs integer programming to formulate the optimal fault tolerance strategy based on the criticalities of different system components. As shown by the experimental results, PFT4MTS outperforms four existing representative approaches in ensuring the quality of quality of multi-tenant SBSs upon runtime anomalies.