Framework for the Distributed Computing of the Application Components

Abstract

Writing real world distributed applications is a challenging task. Even if well known models or powerful frameworks such as MapReduce or HADOOP are employed, the complexity of the aspects involved, such as specific programming and data models, deployment scripts or a hard debugging process are enough to require many working hours or even make the entire process unsuitable for practical purposes. For applications which need some of their own components to be computed in a distributed manner, a generic model incurs an unnecessary overhead and makes the whole development slower. We propose a MapReduce framework which automatically handles all the distributed computing tasks such as computing resources abstraction, code deployment, objects serialization, remote invocations and synchronizations with only a minimal coding overhead. With only minimal constructions dependable distributed components can be developed and run on heterogeneous platforms and networks. The presented results confirm the performance of the proposed method.