A Monotonically Increasing (MI) Algorithm to Estimate Energy Consumption and Execution Time of Processes on a Server

Abstract

It is critical to reduce the electric energy consumption of information systems, especially clusters of servers. In order to reduce the energy consumption, we have to select a virtual machine on an energy-efficient server to perform an application process so that total energy consumption of servers can be reduced. Here, we have to estimate how much energy a server consumes to perform processes. In the simple estimation (SP) algorithm previously proposed, every current process is assumed to have the same computation residue. The execution time of each process and energy consumption of the server are underestimated while it takes short time to do the estimation. In this paper, we newly propose a novel monotonically increasing (MI) algorithm to estimate energy consumption of a host server to perform application processes. Here, current processes $p_1$, $\ldots$, $p_{n_t}$ on a server $s_t$ are assumed to be totally ordered so that the computation residue ${RP}_i$ of a process $p_i$ is ${RP}_1 \cdot {\alpha }^{i-1}$ larger than a process $p_{i-1}$ (i $=$ $1, \ldots , n_t$). In the evaluation, we show the execution time and energy consumption of a server obtained by the MI algorithm are only one [%] different from the simulation results.