An algorithm based on response time and traffic demands to scale containers on a Cloud Computing system

This paper proposes a cloud computing architecture based in containers and on an algorithm that intends to achieve an efficient allocation of processing resources to comply with response time requirements in a Web system. The algorithm is based on the characterization of web requests and on a PID (Proportional - Integral- Derivative) controller. The proposal was evaluated with a real time series obtained from an operational massive web system in a controlled infrastructure. The results show that the proposal achieves the expected response times allocating a lower number of containers than other related proposals.