Analysis and modeling of dynamic capacity provisioning problem for a heterogeneous data center

Many Internet-scale services are emerging and developing quickly recent years, and usually have different performance goals, which is a challenge for a data center to satisfy. Moreover, servers in a data center are usually heterogeneous, which makes it more sophisticated to dynamically provision proper capacity for the data center to adapt to the varying applications workload. In this paper, we first analyze workload data from a publicly available Google cluster trace. Based on analysis results, we model the dynamic capacity provisioning problem in a heterogeneity data center, considering energy consumption and performance together to deal with these Internet-scale services.