Online real-time energy consumption optimization with resistance to server switch jitter for server clusters: Online real-time energy consumption optimization with resistance…: Z. Xiong et al.

Adjusting the deployment of each server in a server cluster in real-time and online based on a changing load has important considerations. In response to the deficiencies in the existing research on server switch jitter and real-time optimization, this paper proposes a periodic energy consumption optimization strategy based on mixed integer programming (MIP) for server clusters. During an optimization period, the strategy allows the CPU of the server to switch between adjacent frequencies to optimize cluster energy consumption at a granular level. First, we describe cluster energy optimization as a basic MIP model, with a reasonable definition of the decision variables and the modeling of the server load and power. Then, we include the server switch overhead in the objective function of the model, considering the joint optimization of multiple periods. Finally, we design an efficient solution scheme based on Gurobi and create two solution adjustment schemes that can reduce CPU frequency switching. The test results reveal that the proposed strategy can effectively suppress server switch jitter and can be carried out in real-time. The extra power cost of reducing CPU frequency switching is also evaluated and analyzed in the testing section. [ABSTRACT FROM AUTHOR]