Model Transformation Engine Implementation Methods Based On MongoDB And Hadoop

Modelling and model transformation is one of the key process in software devel-opment and maintenance.As a special kind of operation,model transformation aims at transform one kind of model into another in a automatic,or semi-automatic way,which is a important component of software development and maintenance.With the growing popularity of computer technology,both the size and complexity of software systems is readily increasing,at the same time giving rise to new challenge for model transfor-mation.As previous model transformation engines are based on stand-alone systems,information kept in memory,the transformation capacity is limited by the physical per-formance the stand-alone system.One the one hand,in the face of complex models,it is prone to excessive memory consumption,so far as to machine crash.Furthermore,the transformed model are stored as files,not suitable for model query.On the other hand,when transforming multiple model simultaneously,previous model transforma-tion engines can only work in a serial mode,with low execution efficiency.Thus,this thesis focuses on how to implement model transformation in distributed storage and parallel computing platform.MongoDB is a NoSQL database based on distributed file system,which adopt schema-free document data model and can store data with different structures flexibly,especially suitable for complex structural models.As no need to read all of the infor-mation into memory when using MongoDB for model transformation,it can effectively reduce memory requirement.Also,the transformed models are stored in MongDB,thus facilitating model query.Furthermore,MongoDB supports distributed data access and can conduct effective collaborative model transformation.?Hadoop is an reliable and efficient open source distributed processing framework,completely encapsulating the underlying implementation details to the users.Using Hadoop as the computing platform for model transformation engine,it is feasible to conduct multiple model transformation in parallel,cutting down the transformation time.Based on distributed storage database MongoDB,and distributed computing plat-form Hadoop,the thesis propose two model transformation engine implementation,mainly including:propose a model transformation engine implementation framework that converts the semantic model transformation into three-tier structure,each encapsulating a set of basic operations(transformation primitives)for model transformation,to lay the foundation for the implementation of model transformation engines.in order to reduce memory consumption when transforming complex models and facilitate model query,propose a implementation of model transformation base on MongoDB,which mainly consists of a model storage algorithm(by which we can store model into MongDB losslessly)and a model transformation algorithm base on a set of primitive operations.in order to take advantages of distributed computing platform to conduct model transformation in parallel,thus speeding up the transformation speed,propose a implementation of model transformation base on Hadoop,which consists of a task decomposition strategy for model transformation(which provides a scheme to de-compose transformation task and execute in parallel using Hadoop),a method to construct MapReduce task base on a set of primitive operations,and a code genera-tion project(which can convert transformation primitives into MapReduce process-ing code).Based on the above work,in order to further verify the feasibility and effectiveness of the method,conduct case study and analysis for each implementation.