Research On Allocating Storage Strategy Based On High-speed Data

Today is an information day. With the development of computer technology and database technology, information technology, more and more widely be applied to all walks of life, people deal with the widespread use of computer technology to the business, coupled with a significant increase in demand for such applications factors can cause the explosive growth of network traffic. Currently, large DBMS in the massive data processing, online business applications has become a constraint database system operating efficiency and even improve the performance of the entire information system bottlenecks, The traditional use of a single database server to improve the hardware configuration database system parameter optimization approach can only be limited to a certain extent, improve system performance. In considering the protection of existing investments to achieve the cost and meet performance requirements, based on parallel processing based on multi-machine database cluster as a cost-effective parallel / distributed computing resources to become a research focus, it has the high performance, high availability and high scalability, it is widely used in mass data processing and storage, high performance computing, e-commerce, web services and other fields.Focusing on improving the performance of parallel database cluster system discussed, the main aspects of the database cluster system structure, middleware structure and design, load balancing and data distribution strategies, etc. to solve the cluster system, some key issues.In order to apply cluster technology into the database system to achieve high parallel processing performance and large-scale database cluster system, a middleware for high performance parallel processing systems based on a database of non-shared storage clustering is designed. providing a single system image structure, use of metadata management, multi-thread mechanism, pretreatment of things in parallel technology, achieve the database cluster system collaboration and parallel processing, to meet the performance requirements of high-line processing (OLAP and OLTP) and other commercial applications. The system is a cluster while maintaining the autonomy of local database sites has improved the performance of parallel cluster database system to solve large-scale DBMS's performance bottlenecks.Access to improved resource utilization and high-performance parallel processing power of the effective means is to use load balancing techniques, the use of feedback mechanisms, the load balancer periodically collection of homogeneous or heterogeneous cluster node imbalances, CPU utilization, memory and network bandwidth and other resources in the comprehensive utilization rate under different loads measured by the weighted load status of nodes, and node performance and running under the appropriate threshold setting.Another important factor which affects the performance of clustered database system is the balance of data distribution, some of the data by the existing method does not take into account differences in heterogeneous computing nodes, and data will be evenly distributed between the various processing nodes to seriously affect the performance of clustered database system. For the system load data skew, the load data migration algorithm used to access features of the data will be migrated to less hot load nodes, not only to avoid the data skew effects and improve the throughput of the system, maintaining the dynamic cluster load-balancing.Finally, experimental analysis and verification of clustered storage systems and parallel processing performance indicates to assess whether the desired effect.