An Implementation of Novel Cloud Computing Model - Improved Map-reduce Model

Although Map-reduce model is easy to deal with querying on a large scale data cluster, but difficult to deal with complex science computing due to its narrow application fields. Among complex applications, most user applications are not querying request simply, but query after complex computing, such as encryptions. This paper proposes a novel method which transforms user´s application task DAG (directed a cyclic graph) into MCST (minimum cost spanning tree) firstly, and then works with map-reduce to complete parallel query. This idea can effectively deal with comprehensive problems which include complex computing and query on a large scale data cluster, such as TSP (travel sales problem), dynamic scheduling, route scheduling and others.