Reliable data distribution middleware for large-scale massive data replication

For massive data replication, Infocast, a communication middleware, is proposed. It distributes large data sets to thousands of receivers without any data loss. This paper examines the distributed processing feature of Infocast and also evaluates its performance. Reliability is achieved by an effective receiver-initiated recovery algorithm and server's complete supervision of all receiver states. Infocast provides reliable data distribution with high-performance to most receivers through the advantage of network-provided multicast while also supporting temporarily unavailable or performance impaired receivers by parallel state transitions. Implementation and performance test results including very large data distribution by VLDD scheme are examined through experiments. Scalability of receiver number and data size is discussed for various conditions.