| Data, which is the precious resource in the information age, has been attached great importance to all people. The immergence of cloud computing service and cloud storage service provide an elastic service for users. It is a novel method to effectively solve the problem of increasing demand of IT storage resources.Compared to the traditional data storage technologies, cloud storage has advantages in low cost, strong scalability, storage management, convenient and high data availability. However, cloud storage system can not promise data reliability for users, the files stored in the cloud have to also carry the risk of data loss. Once the cloud storage service encounters unexpected problems such as disaster and human error, the user files in the cloud will be inaccessible or even lost. Service discruption fault has been happened in the cloud computing systems and cloud storage systems like Amazon and Microsoft. The service discruption not only brings great loss to the users, but also causes commercial trust crisis and the economic loss for cloud service providers. Consequently, the data reliability and data avalibility have become the hot issues of common concern. Cloud service providers have to solve the problem of data reliability in order to obtain better business reputation and economical benefit.Data reliability can not be competely guaranteed in the cloud storage system, thus, the focus of this paper is to build a cloud storage system which can ensure data reliability and implement the optimization of system. To address this issue, this paper designs a multiple heterogeneous cloud based storage system called rich cloud storage system. In rich cloud storage system, each cloud service provider is able to rent the storage resource from other cloud providers. The rich cloud storage system can help cloud providers improve storage resource sharing and interoperation to resist the disaster. The rich cloud storage system has advantages in both of data reliability and cheapness. The detailed advantages of rich cloud storage system is given as follows:●Data reliability:In the rich cloud storage system, user files uploaded to the rich cloud storage system will be copied and stored in the different cloud storage platforms. This mechanism is able to guarantee the Geographical isolation of each data replication in order to improve the data invulnerability.●Scalability:The cloud service provider is capable of dynamically using the storage resource of other cloud platforms to expend the storage capacity.●Data sharing:Data sharing can be easily realized by offering access to other data users.The research content and information of this paper mainly includes two parts:The research on reliable cloud storage system architecture and the research on resources optimization scheduling based on the reliable cloud storage system. The first part focuses on the data reliability of cloud storage system. The research plays an emphasis on the architecture design of the reliable cloud storage system by analyzing the data redundancy scheme, data deduplication and distributed storage, then a novel reliable cloud storage system called rich cloud storage system is build to provide data protection for both cloud providers and users. The resource optimization problem of is taken into consideration in the second part. Both of the cost efficiency and performance improvement are taken into account to optimize the rich cloud storage system. The main contribution of the paper includes:(1) In the research of the key issues in reliable cloud storage system, we make the effort on introducing the technologies of data reliability, cloud storage, and resource scheduling. We first present the research progress of cloud storage technology including virtual storage, data deduplication, distributed storage and load balancing. Then the the recent development of data reliability technology is introduced in detail. We start from the concept of fault tolerance and disaster recovery, and further analyze data redundancy scheme and disaster detection technology of current storage system. Finally, the description of resource management and scheduling strategy in cloud computing system is given in two aspects of system performance and economic benefit. It is the fundamental of the Subsequent chapters(2) The Architecture of Reliable cloud storage system, rich cloud storage system, is designed to protect the user data stored in the cloud. In order to improve the performance and data reliability of current cloud storage system, we proposed a novel storage system called rich cloud storage system, in which the data replicas are distributed across multiple cloud storage providers both private and external. With help of this system, private cloud is able to implement data replicas by utilizing the storage resource from other cloud providers. In rich cloud storage system, data reliability of user data can be guaranteed by store the replicas to different geographical isolation cloud storage platform. In addition, two reprentative applications of disaster recovery and data archival are taken into account in this paper. We respectively design rich cloud based disaster recovery system and rich cloud based data archival storage system in order to provide reliable service for both applications.(3) We proposed a novel Cost Aware High Data Reliability Provision Algorithm (CAHRPA) in order to reduce the cost of disaster recovery for cloud providers. Data availability is a desirable feature for all cloud providers. However, adoption of disaster recovery will bring incremental cost due to the investment of hardware. This paper illustrates the existing assurance mechanisms for data reliability and points out that the data availability and disaster recovery cost are always two incompatible goals for cloud providers. In order to minimize the DR cost at the same to guarantee the data availability, a novel Cost Aware High Data Reliability Provision Algorithm (CAHRPA) is proposed to tackle such an optimization problem. The feasibility and efficiency of this CAHRPA is verified by the comparison with some other replication strategies.(4) Data archival and de-duplication have become commodity components for cloud storage platforms. In this paper, we present Rich Cloud Based Data Archiva System (RCBDAS) that practically addresses the reliability of cloud storage. Due to that our RCBDAS is a complex system consists of different heterogeneous cloud storage systems, the performance of data transmission depends on the bandwidths of different cloud platforms. In roder to improve the network performance of RCBDAS, we proposed a Network-Aware Reliability Provision Scheme (NARPS) including data deduplication, data distribution, repository selection and redundancy placement. With help of NARPS, RCBDAS not only achieves fault tolerance of storage, but also allows network-aware data redundancy strategy. Finally, we build a working implementation and run some benchmarks in our RCBDAS. We can conclude from these experiments that RCBDAS enables clients to explore a network-aware data redundancy scheme by using multiple heterogeneous clouds.In conclusion, the aim of this paper is to develop the data reliability mechanisam of cloud storage system. The research emphasis of our study is how to bulid the reliable cloud storage system and how to improve the system performance. In order to guarantee the data availabitliy of storage system, we proposed a novel rich cloud storage system, in which data are distributed to different heterogeneous cloud storage platforms. In particular, two specific systems are presented according to two different storage applications. In addition, the economic benefit and system performance are taken into consideration and two optimization schemes are proposed to solve these problems. Finally, we demonstrate the effectiveness of our proposed scheme on end-user tasks. |
PDF Full Text Request