DCFMS: Data Handling Distributed System for Achieving High Data Availability

High data availability is crucial in every high performance environment. Even though network environments have improved a lot, costly data transfers remain the main drawback in achieving high performance in complex distributed applications which need to come up with additional design techniques to overcome this drawback. Continuing our previous research which focused on the design and implementation of a computational steering enabled distributed framework, we have developed a data flow management system that is able to isolate the data handling from the computational environment. The system does not focus on defining new transfer techniques for the physical environment, instead the innovation comes from two main techniques that avoid costly transfers: logical partitioning and data localization. The experimental evaluation of the system is also included.