Research And Implementation Of Elastic Load Balancing Mechanism On Service Innovation Platform

Some of the applications deployed on the Service Innovation Platform are under intense operational pressure in actual operations. At the same time, these applications themselves also have a requirement for testing and validating high concurrent traffic. Therefore, the platform needs to provide these applications with running guarantee and experimental environment by the load balancing service.The applications need to specify the number of load machine before the load in the traditional static load balancing mechanism. It is not suitable for those applications whose business volume has great fluctuation. The load balancing service on the service innovation platform should be more flexible and the applications could do the real-time adjustment of load resources on the basis of the degree of business pressure. This kind of service is elastic load balancing mechanism. The elastic load balancing mechanism can adjust the load machine dynamically according to the monitoring feedback information, but this kind of adjustment is passive and delay. The service innovation platform should reduce the lag of time as much as possible. The operation and maintenance personnel of the applications need to work on complicated configuration while performing load balancing configuration. The service innovation platform expects to automate this process and maximize efficiency for the staff.In order to achieve the above objectives, this thesis studies and implements an automated intelligent elastic load balancing mechanism on the service innovation platform. First, the elastic load balancing service is implemented on the platform and the system can dynamically adjust the application's load resources based on monitoring feedback. It involves the application of Docker technology, the deployment of the underlying network, OpenStack source code modifications, Libvirt monitoring, etc.Second, the prediction mechanisms and the resource pooling module are implemented to solve the problem of hysteresis in elastic load process. It involves decision tree classification algorithm, history information storage and so on. Last, the front-end asynchronous interaction module is implemented to provide the load balancing management interface for the application operation and maintenance personnel. In the last part of the paper, several experiments and tests about the intelligent elastic load balancing mechanism are designed, which verify that the proposed mechanism can effectively operate on the cloud platform. Experiments show that the use of the mechanism has greatly improved the ability to cope with peak traffic and average response time is decreased by about 34%.