Abstract:
Over the past decade, the energy demand has been increased rapidly. The rapid surge in
demand for electricity due to the exponential growth of population has been considered as one of the
most significant challenges faced by the modern world. As a result of their massive energy needs
and high computing demands, large-scale data centers account for a sizeable share of the costs
connected with running a data center. Consequently, the financial cost of energy use and the carbon
emissions caused by energy use are becoming serious issues for society. This enormous energy
demand for data centers and related businesses may eventually reach a bottleneck. Several sorts of
research have been performed to build energy management frameworks with the goal of managing
energy demand in data centers through the integration of renewable energy sources and energy
storage systems. However, it is difficult to rely on renewable energy because its supply is based on
a variety of circumstances. Though previous studies attempted to more precisely anticipate the
amount of generated renewable energy, adequate renewable energy supply cannot always be assured
due to variability in its affecting factors (for example, climate, time of day, temperature, etc.). As a
result, a technique must be developed to ensure the supply of renewable energy despite its
intermittent nature.
In this research work, we have proposed an energy management framework for data centers by
a combination of solar & grid energy. A service-level-objective (SLO) will be assigned to every
individual zone of data center, that will be specified by each data centers’ energy requirement. Jobs
with the same SLO will be assigned to the same data center zone, and each zone will be fueled by
renewable generators with a chance of generating electricity equal to or greater than the area's need.
To address the intermittency issues associated with renewable energy sources and reduce SLO
violations, an effective mapping strategy of renewable energy sources and data center zones will be
developed using the reinforcement learning technique Deep Q-Network (DQN) Algorithm to ensure
maximum data center uptime.
