Cloud computing requires further research and development to accommodate more application areas . We introduce a new application area: industrial automation. A current industrial automation (IA) system is a multi-tiered architecture entailing different layers from feedback control to enterprise management. If adopted in large-scale IA systems, cloud computing can offer over 40% cost saving and 25-85% time saving [4, 1]. However, IA requires tighter timeliness, reliability, and security than most other cloud applications. We propose a cloud-based IA architecture and focus on the timeliness and reliability requirements. Addressing such requirements for the lowest layer (feedback control) is the most challenging. We addressed the timeliness problem in . To address reliability and further address timeliness, we propose a distributed fault tolerance algorithm for cloud-based controllers. We theoretically and practically prove that the proposed fault-tolerant, cloud-based controllers offer the same performance of the local ones.