The application of P2P in data warehouses

摘要

In this work we have described a possible approach to integrate peer-to-peer support into a HDFS architecture by using a Distributed Location Service as a means to connect cluster machines and allow communication among them in dynamic scenarios, thus removing the need for ad-hoc configuration of the HDFS system. The peer-to-peer layer underlying the HDFS system is meant to provide a hot-plug framework for data nodes to dynamically connect and disconnect from an HDFS working session without compromising overall system's performance.