Efficient and Reliable Execution of Legacy Codes Exposed as Services

In this paper, we propose a framework that enables fault tolerance and dynamic load balancing for legacy codes running as backends of services. The framework architecture is divided into two layers. The upper layer contains the service interfaces and additional management services, while the legacy backends run in the lower layer. The management layer can record the invocation history or save state of a legacy worker job that runs in the lower layer. Based on this, computing can be migrated to one of a pool of legacy worker jobs. Fault-tolerance in the upper layer is also handled by means of active replication. We argue that the combination of these two methods provides a comprehensive support for efficient and reliable execution of legacy codes. After presenting the architecture and basic scenarios for fault tolerance and load balancing, we conclude with performance evaluation of our framework.