The Research Of Quality Of Service And Reliability For Cloud Storage System

The rise and development of cloud computing technology is gradually changing the way to use services and resources, and also being a hot research field of computer applications. Cloud storage can store the mass data efficiently, and is an important basis for cloud computing. How to ensure the quality of services (QoS) and reliability for cloud storage system, is of great significance to guarantee the efficiency of accessing data for various users and maintain the availability of mass data storage. Now the mainstream cloud storage systems, which are lack of effective QoS control mechanisms, can not meet the needs of different users to access data; the other hand, the static protection technology for the reliability of cloud storage system is difficult to adapt to the changing system environments. Thus, at the basic of national science and technology projects with research, and the structure of cloud storage system model, the paper does research for the QoS control and reliability technology for cloud storage system.To ensure the QoS of data access for Cloud data storage system, the paper proposes a DiffServ based QoS control method which provides different QoS guarantees for different users based on the different quality of service levels. The one hand, the metadata server that using EDF queue scheduling meets the needs of responding delay; in data storage server, it uses multi-level queue scheduling methods to meet the transmission bandwidth of the output, and ensures the QoS of high-level user firstly through adaptive load regulation. On the other hand, we use the QoS-aware load balancing strategy for regulating the QoS control of the whole system, to implement the QoS optimization for the allocation of read request. Finally, using the HTTP mechanism ensures the QoS of data transfer process.Beause the data storage serves are not reliable, the paper proposes a bi-directional failure detection strategy, which is composed of heartbeat detection strategies and active detection. In the heartbeat detection, we use the adaptive heartbeat timeout algorithm to adapt to the changing cloud storage network environment; when a heartbeat is time-out the active strategy is triggered to avoid the miscarriage of justice caused by packet loss. Experiments show that the bi-directional failure detection system is well adapted to the changing environments, reduces the detection error rate and accurately identifies the down data storage server.To ensure data reliability, the paper presents the initial placement strategy, adaptive control strategy for the number of replica and replica-failure handling mechanism which we call them the reliability-oriented replica strategy. T^e initial placement strategy selects the best server for placing the root replica to ensure the reliability of the initial storage. Based on the user's access, adaptive control strategy for the number of replica adjusts the number of replica dynamically, increasing copies and reducing the load of the hot files; it also reduces the unpopular copy and saves the memory resources. When a single replica is broken or a data storage server fails, failure handling mechanism will start and ensure the number of copies by quickly copying a replica of the redundancy, to guarantee the reliability of access to data.At last, the paper introduces the prototype system which bases on the QoS control and reliability technology:JTang Cloud storage System. JtangCS achieves the QoS control and reliability technology described in our paper.