The Study Of A Mechanism Of Self-deploying For Large Applications

Software deployment is an important stage of software life cycle. It not only affects the performance of software, but also affects the efficiency of runtime environment. Now days, for large monolithic applications, software engineer usually use the deployment strategy of installing large monolithic applications on a single server. This deployment strategy demands high configuration of servers, and demands high input costs. What's more, this simple deployment strategy does not suit for the environment of pervasive computing and mobile computing. However, deploying a large monolithic application into open, dynamic and heterogeneous network environment is a very complex process, which not only needs to consider structure of applications, models partition and models communication, but also needs to consider resource constraints of network environment and user's requirements.According to the principle of autonomic computing, this thesis proposed a self-deploying framework for large monolithic applications. Extracting weighted class graph using static analysis of applications, using clustering technology to analysis weighted class graph, partitioning application into more components according to clustering results, using mathematical programming to deploy more components into complex network environment finally.Experimental analysis proved that the hybrid clustering algorithm performs best in terms of clustering quality, authoritativeness and extremity of cluster distribution, which can help to partition large monolithic applications into more components. Mathematical programming is easy to describe the resource constraints of network environment and user's requirements. It performs better in the process of software deployment.