基于贝叶斯网的移动互联网应用部署算法研究

针对传统的应用部署算法多是对特定平台上的简单应用提出,不适合对复杂应用实现有效的部署.提出了服务子图的概念,简化了应用拓扑结构,降低了应用模型的复杂度,更加适用于大型移动应用;另一方面,采用基于贝叶斯网络及概率图的算法来维护应用部署预测模型,提高了应用部署的性能和准确性,提高了服务质量;最后搭建了实验平台并研发了典型应用进行验证,取得了较好的效果.

Deployment strategies based on Bayesian networksfor mobile application

The reasonable deployment strategies are effective ways to improve the quality of mobile application services. How to improve the partition efficiency is the research focusin terms of mobile Internet applications. Most current researches use static algorithms to clarify the simple application at a specific platform,and fail to apply them to large-scale and complicated mobile applications. This paper proposes an application partition model andan algorithm. At first,the sub-graph theory is used for simplifying the application of the topological structure and reducing the complicity of the model. For meeting the accuracy and real-time requirements,this research uses the Bayesian classifier to implement the static partition algorithm for its application deployment. At last the research builds a mobile platform and develops several classic applications to verify these concepts.