Contents I. Introduction
The emergence of data centers powered by renewable energy consists of a new research area. Different sources of renewable energy daily draw energy from natural resources and that energy are stored in battery storage systems. The batteries revert the electrical power into a chemical reaction that can be stored and reversed when necessary. However, as a physical device, the cells are limited in the amount of storage. The data center classified as Tier 2, 3 and 4 shall provide, respectively, redundant components without stopping maintenance, and fault tolerance by providing more than 99.749% availability, with a maximum downtime of 22 hours [1]. Among the requirements required for such tiers, one consists of multiple power inputs from different power sources and operators. When renewable energy is used, there are two basic ways of utilizing the generated energy: (i) energy is used in real-time, and the surplus is " sold" to energy operators as a discount on general power consumption, or (ii) The energy is stored in batteries, and allows its use within a predetermined period, thus allowing the batteries to be charged again in the next cycle. This work is located in a context where the data center has more than one power input, one of which is powered by renewable energy, and all energy absorbed by renewable energy sources is stored daily in batteries. For this reason, models that can address daily power budgets [2] become necessary in these environments because the energy stored in the batteries each day can only support a slice of the total resources of data centers. It occurs because the amount of renewable energy produced is still far short based on the needs of large data centers regarding power supply. Because of this, most data centers use hybrid systems, being a part supported by renewable sources and the majority of conventional energies. From the above, this work proposes a combinatorial optimization model to several computational resources compatible with the daily power budget of different available renewable energy sources. This proposal would maintain sufficient resources based on the offered green power. In this sense, the model proposes to provide a sustainable data center, which reduces its impact on sustainability issues. Based on this purpose, this paper presents the following contributions: (i) A review of the types of renewable energies that can maintain large-scale computing environments such as data centers, besides the determination of a daily power budget for each one, (ii) A combinatorial optimization model for the selection of the highest amount of computational power, obeying a daily power budget of each type of renewable energy, and (iii) Highlights and trends for sustainable data centers, based on the results presented in this work.
