Cloud Computing-based Integratedoperation Management Platform Research

In recent years,internet information technology and SOA technology increasingly mature,and increasingly popular in the field of practice.In the current international research field,the new generation of cloud computing technology based on the concept of service has become a research hotspot of experts.There is no doubt that under the impetus of cloud computing,information technology is developing towards intensification and specialization.Cloud computing makes the links between the information technology industry chains increasingly close,and the effect of resource aggregation is increasingly prominent,which further promotes collaborative work and information sharing,and provides a new way of thinking for the service oriented computing era.Aimed at the problems of the integrated management system for operation and maintenance “information island” and the flexibility of management system's functionalities,a framework of the network management system for operation and maintenance based on cloud computing is introduced.It assimilates the thoughts of virtualization and SOA,designs a cloud computing platform encapsulating the network management functionalities to loosely coupled services,and breaks the barriers of original systems.The following issues about distributed operation service selection based on cloud computing,the method of operation data classification based on Hadoop and the clustering algorithm based on Hadoop are proposed as the focus of this dissertation.In order to deal with the unpredictability caused by the randomness of the cloud,an online algorithm based on learning method is proposed.In this paper,we consider the online provider selection framework,where users dynamically and individually select their service providers based on experienced performance,and investigate the distributed decision-making strategy to achieve overall and individual performance guarantee.Specifically,we propose the learning-based selection policy,named Exp3.C,which regulates the system converging to a set of pure Nash equilibriums(PNE)of a congestion game in the homogeneous scenarios.Further,we show that even in a chaotic scenario where cloud users maybe irrational(which results in disordered and unpredictable behaviors)and the available resource of providers may change,the user's profit is guaranteed to approach that of selecting the best provider.Finally,simulation results show that the stability of the algorithm can improve user satisfaction and guarantee service quality according to user sensitivity.In order to optimize the SPRINT algorithm,the calculation method of the best segmentation point is improved.Through two new data structures,the type partition table and the merged partition table make the operation process more streamlined,so that the unnecessary operation can be reduced to reduce the number of candidate nodes.In this way,the time needed to construct the decision tree is shortened,and the performance of the algorithm is optimized.Aiming at the requirement of classification and processing of massive data,a data classification implementation model based on Hadoop platform is proposed by combining data classification technology with cloud computing technology.Considering model requirements,model composition and hierarchy structure,the data classification is combined with the Hadoop framework,and the SPRINT algorithm is improved.In particular,the optimization work is carried out from the aspects of sort parallelization,node parallelization and attributes parallelization to ensure that the improved algorithm can be adapted to the Hadoop platform.In view of the slow convergence of ant colony clustering algorithm,according to the principle and the feature of convergence with the previous faults at slow speed.We introduce the principle of data preprocessing.Through K-means algorithm for data preprocessing,we reduce the time cost of the algorithm.It assimilates the thoughts of virtualization and SOA,designs a cloud computing platform,encapsulates the operation and maintenance management functionalities to loosely coupled services,breaks the barriers of original systems.Then based on this framework,the operation and maintenance management framework is implemented.The framework can meet the need for virtualization,service-oriented,integration and reusability.Aimed at the problems of the network management system for operation and maintenance “information island” and the flexibility of management system's functionalities,a framework of the network management system for operation and maintenance based on cloud computing and SOA is introduced.Firstly,the task oriented cloud computing service and data resource management architecture are studied and its components are introduced.Secondly,the cloud computing network resource monitoring technology based on management agent is designed,and its operation mechanism is also described.Finally,the adaptive cloud computing network planning and scheduling technology to task are introduced,and the resource view and description method of its requirements are analyzed,and a virtual machine network planning method based on AHP is proposed.