Research And Implementation Of Cloudstack AutoScale Based On KVM

AutoScale technique can help cloud computing service providers improve hardware resources utilization and customer experience. In addition, it can also reduce costs for users and make cloud resources achieve the desired purposes. With the rapid development of KVM virtualization technology, current AutoScale technique developed by CloudStack community, which is based on XenServer, has been unable to meet the needs of technology and market. It exists the problems of unable to support the current popular open source KVM virtual machine monitoring program, poor monitoring algorithm performance, high response delay and small system throughput. Moreover, its corresponding basic network load balancing algorithm is too simple to achieve the purpose of load balancing, which leads to the frequent occurrence of load imbalance. How to effectively solve the above problems is one of the hot spots in the research field.This thesis launches several researches on the AutoScale schemes of KVM virtual machine monitoring program based on the above background. The main contents are as follows:1. Proposed the AutoScale monitoring algorithm which based on multiple commands. The AutoScale monitor algorithm based on XenServer has the disadvantages of high response delay, poor monitoring data consistency and unable to allocate and recycle hardware resources rapidly according to the load condition of virtual machine, which results in idleness and waste of hardware resources. The AutoScale monitor algorithm based on multiple commands proposed in this thesis has solved this problem effectively. The experimental results reveal that compared with the control algorithm based on single command, the algorithm based on multiple commands significantly shortens response time of monitoring system and improves consistency of monitoring data.2. The research puts forward a weighted round robin load balancing algorithm model based on dynamic feedback. The implementation of AutoScale scheme needs the support of load balancing. However, the current CloudStack load balancing algorithm has poor performance, small system throughput and long response time, which can't achieve the purpose of load balancing. The weighted round robin load balancing algorithm proposed in this thesis effectively improves the throughput of system and reduces the load imbalance phenomenon. The experimental results of three different load cases show that this algorithm performs better on response time and throughput than traditional round-robin algorithm and weighted round-robin algorithm.3. The AutoScale function module based on KVM has been finished. The current CloudStack technique can't support KVM virtual machine monitoring program, and its high response delay and small system throughput can easily lead to load imbalance. In this thesis, the design and develop work of AutoScale architecture and function module are achieved after combining the interval virtual machine monitor strategy, monitor algorithm based on multiple commands and weighted round robin load balancing algorithm based on dynamic feedback together. System running results verify the feasibility and effectiveness of this model and algorithm.Finally, the module design and development parts have passed relative function and performance tests and been applied to the enterprise private cloud production environment, meeting business public and private cloud needs. Moreover, some modules have been deployed in the enterprise core software, aroused wide praise among users, and will be gradually applied on the public cloud platform.