Modeling Energy Savings in Volunteers Clouds

In this paper we propose different models for the energy consumption in a special existing cloud system named SlapOS. In this cloud, the data center comprises dedicated and volunteer machines, these latter ones are not always available. Our objective is to state how to plan the run of applications for minimizing the global energy consumption, we propose two modelings. In the first model, we assume that we have a finite number of homogeneous volunteers nodes on which our applications can be run. The objective is to determine which application to run on each node in order to minimize the overhead in energy consumption caused by these runs. We show that the key computational challenge in this problem consists in finding a feasible solution when it exists. We propose for it a polynomial time algorithm. In the second model, we assume that the volunteers nodes are heterogeneous. In this case, we show again how to find a feasible solution in polynomial time. But in comparison to the homogeneous case, the key computational problem to solve here is NP-hard. We then propose an ILP (Integer Linear Programming) formulation for addressing and evaluate it throughout various simulations of the SlapOS system in a realistic volunteer computing context.