Coordinated Access to Shared Data Sources for Geo-replicated State Machines

State machine replication techniques are widely used to create fault tolerant services in cloud computing. However, reaching consensus between remote regions may take substantially long time. Many algorithms are proposed to solve this problem regarding the geographic locality of the objects while trading off some consistency properties of the system. Most of these studies consider direct replication of arbitrary parts of the system, but contemporary online services share some unique data sources that may not be replicated. This paper proposes a coordinated method to manage shared data sources between replicas by utilizing geographic locality of the data sources. The proposed algorithm present better performance than using a leader based approach in terms of request processing throughput. We provide the design and implementation of the proposed algorithm in detail and present the throughput performance of the algorithm in a geographically distributed environment.