Study on proactive auto scaling for instance through the prediction of network traffic on the container environment

In this paper, we propose container traffic analyzer (COTA) structure to improve accommodating more network traffic to VMs and to reduce the scale-out time. COTA consists of two functions. The one is reporting the amount of network traffic on real-time. The other function is managing server balance on user requests and scale-out VM by using aggregated network traffic profile. Based on these network traffic information, we propose Least Traffic Load Balancing(LTLB) algorithm to solve network traffic imbalance problem. LTLB algorithm establishes new connection to VM which has the least traffic in real-time. We test performance comparison evaluation with existing well-known dynamic load balancing algorithms. And we apply the algorithm to the Docker based Container environment that has light-weight and occupy low storage capacity to provide fast and elasticity scale-out as well as scaling policy including traffic threshold. Then, performance evaluation is done between VM over hypervisor and Docket based Container.