Research On Load-balancing Mechanism Based On SLA In Cloud Environment

Cloud computing service providers by signed SLA (Service Level Agreement) with customer to provide Non-local network resources in a more targeted, higher-speed, lower-cost way, and also And to ensure their own interests. Due to the Large-scale and heterogeneous properties of cloud computing datacenter which will be a serious threat to the server performance, users will not satisfied and service provider needs to pay liquidated damages according to the SLA content. Therefore, how to build an effective load babancing mechanism and reasonable use of network resouces to ensure SLA has became the main contents my thesis.First, we analysis the load balancing differences between physical machine (PM) layer and virtual machine (VM), and compare the granularity of the process migration and virtual machines migration. Focuses on the characteristics of load balancing in the cloud computing environment, based on the fact that the granularity of the process migration is much smaller than virtual machines migration, we established a three-layer load balancing model which includes job scheduling, process migration, vitual machine layers. Based on the above three-layer model we proposed a load evaluation mathematical model according to hardware indicators and real-time resource utilization. Through the load evaluation, we can determine the load state of each physical and virtual machines in the server. The evaluation results provide the basis for the migtation strategy. And considering the computing capability and resource utilization of each receiver nodes we using the forward probability method to make the high-performance-light load nodes are able to give priority to recive the system load.Secondly, taking into account the majority of network applications has the characteristics of change over time cycle, and may appear instantaneous peak load. If temerity launch the migration will result tha the benefit of migration is lower than the cost of it. So we proposed a load balancing launch strategy which has the function of load prediction and SLA-based benefits estimate. The SLA based migration benefit estimation model is used to analyze the benefit and cost of migtarion, only to meet the estimated function, can ensure that migration is beneficial to the entire system.At last, through simulation experiments, verified that the load balancing strategy in this thesis can balance the system load, improve service quality and guarantee SLAs.