Cloud environments at data centers in general have well established architectures for uninterrupted power, security, expansion opportunities and availability. However, the architectures are generally consuming much more energy than required. The problem statement is addressed by developing a prototype to demonstrate how energy efficiency can be optimized. This is done by using three multi dimensional bin packing algorithms. The CERES project addresses both the algorithms for scaling physical compute nodes in a cloud environment and the practical use of algorithms from a system administration perspective. Our results and experiences show that each algorithm has different effect on the balance between energy efficiency and performance. The simplest bin packing algorithm provides useful mechanism to reduce power consumption, while more complex algorithms do not give the same power saving, but give better performance. The project was able to optimize the energy efficiency by lowering the energy consumption with 40% without reducing the performance. An even higher reduction can be achieved by overbooking the hardware capacity given to the virtual machines. Some experiments did reduce the consumption with as much as 72%.