In the enterprise, there are a lot of information systems that built by various vendors on heterogeneous platforms. Due to the lack of unified planning and standards, data sharing and exchanging between them is really a troublesome work. Therefore, how to establish uniform, standard interface to archive data exchange and sharing within distribution, independent, heterogeneous systems is becoming a hot issue.

Data exchange and sharing platform is a solution for this problem, it is based on the unified middleware platform, realizing heterogeneous data exchanging and sharing by providing client API and front lead switching nodes deployed in the application system. In a majority of cases, data exchange platform is star topology deployed and the data type varied acutely from the different application domain. In this scenarios, a large number of nodes need to exchange huge volume data with central node directly or indirectly, this pose a significant challenge to the central node’s throughput and reliability.

To improve the throughput and reliability, this paper proposes a HDFS based data exchange and sharing architecture. In this model, data exchange process is broken down into the meta-data exchange and data exchange two processes, this mechanism can broke data exchange requests to the storage cluster which provide distributed and reliable data storage. Meanwhile, to acclimatize the application scenarios, this paper propose a heat based replica management model, which dynamically adjust the number of hot replica to reduce hot data exchange time; a small file index optimization to improve the efficiency of small file storage; a reliable data exchange mechanism that handle the client and server-side failure.