Research Of Multi-Tenant Technology And Resource Managemnt In Cloud Computing

Cloud Computing becomes the primary source of computing power, since it can be used in a cost-efficient and flexible way. With broad utilization of Infrastructure as a Service, the cost of building a cluster of infrastructure largely decreases and the utilization of infrastructure resources obviously increases. Resource management and resource configuration are two significant problems in Cloud Computing. To solve them, we design and implement multi-tenant service template system and dynamic resource management and monitoring system using Naive Bayes Classifier.In order to customize the vitural machine instance and image file with the tenants’ requirements and share them among tenants, we design and implement the multi-tenant service template system. This system leverages meta-data driving multi-instance multi-tenant, which can be divided into four layers:meta-data layer, API layer, logic layer and Web layer. The meta-data layer extends the OpenStack’s original meta-data schema to build the fundamental data structure. The API layer is responsible for implementation of customization and share of vitural machine instance and image file by using RESTful Web API. The logic layer is in charge of logical control, based on the meta-data layer and the API layer. The Web layer, deriving from OpenStack Dashboard, provides user a concise Web interface.Resource management plays a critical role in infrastructure cloud platform, aiming at avoiding both resource under-utilization and over-utilization. To improve resource utilization, we design and implement dynamic resource management and monitoring system using Naive Bayes Classifier, which is composed of two sub-systems including resource monitoring sub-system and resource management sub-system. The resource monitoring sub-system we implementing with CORBA can be used in the distributed heterogeneous computer architecture. And it can monitor both servers’ and vitual machines’ resources of CPU, memory and network. The resource management sub-system employs Naive Bayes Classifier, which provides dynamic resource adjusting threshold according to monitoring data. And by using vitual machine instance’s dynamic resizing and live migration, the resource management sub-system can realize dynamic resource management to maximize utilization of resources.At last, we conduct some experiments and tests on our system. The results present that the multi-tenant service template system successfully implements customization and share of vitural machine instances and image files. At the same time, the dynamic resource management and monitoring system do a better performance in managing resource in our Cloud Computing platform.