Research On Memory Resource Sharing In Cloud Computing Environment

As development of computing and network technology, internet computing technology continued to make progress in recent years. The business and academic pay highly attention to cloud computing which is a new representative of the internet computing model in just a few years. Memory resource sharing in a general cluster environment has been deeply researched and practicality deployed for application. However, cloud computing data center as a new type of cluster environment, brought some challenge to the management and effective sharing of memory resource because of the imbalance of system load, normal failure of nodes, and other new features during operation.Traditional network memory approaches are usually designed for small-scale cluster system and the architecture adopted of memory resource sharing can not adapt to the cloud computing environment. In this thesis, according to the features of memory resource in the cloud computing environment, we propose a fully interconnection structure based on centralized sub-systems by using traditional network memory approaches. Memory resource is managed in the structure through a Super master and a number of Master nodes. We also propose a distributed service structure of memory resource to manage all the Slave nodes in a subsystem. Each Slave node in the service structure is an independent network memory system. Slave node meets different capacities of memory resource demanded by users through two memory service mechanisms—multi-service principle and dual identity principle. Experimental results show that the fully interconnection structure based on centralized sub-systems can effectively manage memory resource in the cloud computing environment and solve the problem of load balancing, and the memory resource sharing system has good performance.Due to the special features of memory resource, it can hardly borrow the traditional centralized, unstructured decentralized and structured decentralized resource information management systems. Under the fully interconnection structure based on centralized sub-systems, nodes are grouped by the rack in the cloud computing data center, and each rack is a centralized resource information management subsystem. The system load is not balanced in data center. According to the feature of data center, we propose a service group model and corresponding self-adapting aggregation algorithms. The algorithms dynamic manage the memory resource of rack nodes based on the demand of users. Experiments show that our proposed methods not only can efficiently manage the memory resource information but also improve the overall performance of memory resource sharing systems.In the cloud computing environment, node failure is the norm rather than the exception, and it brings challenges to the reliability of memory resource sharing approaches. According to the characteristics of memory resource in the cloud computing environment, we propose a structure of reliability service for memory resource, and there are a backup nodes selection mechanism, a copy consistency protection mechanism and a service switch mechanism to ensure the reliability of memory resource for users. Simulation tests show that the proposed guarantee mechanisms of service reliability for memory resource ensure the reliability of systems.Key Words:cloud computing, network memory, memory resource sharing, service group module, self-adapting aggregation, service switch...