Research On Resource Adaptive Equilibrium Mechanism Of Multi-granularity Load And Its Application

As the most basic aspect of application system, DBMS’s performances become critical application performance. Whether large or small applications both systems exist to fully utilize server resources. And now the actual situation is the server’s utilization rate resulted in a waste of resources, but because of lack of resources but also makes effective use of the system’s service quality degradation, the specific performance of the system is unable to respond to the needs of users in a timely manner. For these reasons this paper, the static characteristics of the database load and load characteristics of the probability distribution of the time and research and analysis to better planning database systems and database cluster in the distribution of data in order to achieve the purpose of improving system performance. While different according to each application system load characteristics for the planning and deployment of application systems in order to make full use of each server resources. This achieves the software architecture and hardware resources of the match.In-depth analysis of software performance testing and measurement methods based on the database workload characteristics into the main line, combined load scheduling and balanced strategy to enter the in-depth study of load modeling and database server resources modeling, design and realization In this paper, the modeling method can support the database automatically test platform. Specific innovation for database workload performance and contribution are as follows:●The study the relationship between the database server performance and resource consumption patterns and load. Methods and indicators for load characterization. Quantitative correlation study of the relationship between the performance of the server in this abandoned little impact on the performance of the database load parameters, reducing the the server modeling of performance indicators, reduce the load balancing algorithm complexity and improve the efficiency of the load balancing algorithm. ●Database server tasks based on the load balancing strategy and scheduling scheme. Task scheduling strategy based on the combination of the movement of the load characteristic information. The scheduler combines the advantages of the load balancing, scheduling dynamic and static scheduling scheme, and accurately select the characteristic parameters can improve the accuracy of the load balancing algorithm and the corresponding times.●Builded and realized the operating characteristics of the data layer analysis platform. Database application system, the capacity of the database server, database structure, and storage structure and environment configuration directly affect the performance of the data layer processing tasks. By auto-correlation analysis and conversion process in accordance with the database run log and trace information, in accordance with the load spatial and temporal characteristics or special configuration requirements generate data layer tasking. Script, the script can be executed repeatedly database platform and access to the database server data processing different types of tasks are the resource consumption of the corresponding mode and the processing performance of the server for the different types of tasks.●According to the load characteristic information modeling, using layered queuing network modeling techniques in the operating characteristics of the data layer analysis platform to create a performance model of the database server, the software and hardware of the management server CPU, disk, query agent process, data management process resources.The above study provide the basis for cloud computing platform to achieve optimal allocation of resources and computing power in accordance with the demand to provide a basis for capacity planning of the database server, the storage structure design, the key features of the design and performance optimization. Improve server utilization and achieve the maximum throughput of the service cluster unit time.