Fault Tolerance of Data Center under Multi-Correlated Failures

Abstract

Multi-correlated failures lead to severe power outages in an inter-connected power network. The critical loads of the data center, such as servers and chillers are highly dependent on the grid power. The service reliability and fault tolerance of the data center becomes challenging under multi-correlated failures. Data miss-management and service unavailability under such scenario will result in the massive revenue loss. To address this problem under multi-correlated failures, a fault tolerant model is proposed to minimize the revenue loss. Moreover, considering the system dynamics, our model maximizes throughput and operational time of the data center. Furthermore, we formulate a stochastic optimized system model that captures power consumption, power distribution, data center generation, data center workload, service deadline, and revenue loss. The effectiveness of the proposed fault-tolerant model is tested and validated using real-time data.