Self-Configuration Framework Based On LQNM For Database In Cloud

With the rapid development of cloud computing technology, applications deployed in the cloud computing platform become diverse and complex, so technologies requirements for cloud computing become increasingly high. As cloud computing platform’s backend support system, the performance of cloud database directly determines the performance of applications deployed in the cloud computing environment. one of the effective means that satisfy users’ service quality and system performance goals is adjusting the system resource allocation according to the change of the load at different time. However, faced with the huge amount of data under cloud database systems, complex load, to guarantee the service quality of cloud database system by adjusting the system of resource allocation manually is very difficult, not only requires a lot of cloud database professionals, but this work is very time-consuming and can not meet the real-time performance of the target system. Therefore, building a load adaption configuration framework for cloud database has become a core mission to solve the problem.Adaption configuration framework proposed in this paper is based on the performance prediction model, building a cloud database system performance model to predict the performance of cloud database systems can predict the system’s performance in the case of a given resource allocation by using the perfomiance parameters, thus, searching for the system’s optimal resources allocation own a strong support. Currently, there are two mature performance model:queuing network model and Layered queuing network model. Queuing network model usually apply to simple systems (processing workload request only involves the hardware resources of the system), layered queuing network model apply to complex systems, the process of handling the workload request can be described the relationship between software resources, between hardware resources, between hardware and software resource. Based on the facts above, this paper proposed Layered queuing network model for HBase by analyzing the working process of cloud database. The model consists of five tasks:two software task (work process, database management process), and three hardware task (THINK, CPU, hard disk).In this paper, the experimental environment is HBase as cloud database,1G TPC-H benchmark data as a data source and workload. In the experiment by comparing the adaption configuration framework proposed in this paper with no control mechanism and the priority control mechanism find that system QoS under the control of adaption configuration framework has been greatly improved and prove the effectiveness of adaptive configuration framework. In addition, in order to prove the validity of performance models (Layered queuing network model), by comparing the average response time and the average throughput by using the layered queuing network model to predict with the predicted parameters by queuing network model and the actual measured values, the experiment results find that the predicted performance values by layered queuing network model are closer to the actual performance value than the predicted value of queuing network models, illustrate the accuracy and effectiveness of the Layered Queuing model in predicting the performance of HBase.