No abstract

OATAO is an open access repository that collects the work of Toulouse researchers and makes it freely available over the web where possible.

OATAO is an open access repository that collects the work of Toulouse researchers and makes it freely available over the web where possible. This is an author-deposited version published in : http://oatao.univ-toulouse.fr/ Eprints ID : 15265The contribution was presented at E2DC 2014:http://www.dc4cities.eu/e2dc.html Abstract. In this paper we present an approach to improve power and cooling capacity management in a data center by taking into account knowledge about applications and workloads. We apply power capping techniques and proper cooling infrastructure configuration to achieve savings in energy and costs. To estimate values of a total energy consumption and costs we simulate both IT software/hardware and cooling infrastructure at once using the CoolEmAll SVD Toolkit. We also investigated the use of power capping to adjust data center operation to variable power supply and pricing. By better adjusting cooling infrastructure to specific types of workloads, we were able to find a configuration which resulted in energy, OPEX and CAPEX savings in the range of 4-25%.

CoolEmAll project aims at improving energyefficiency of data centers. The main results of CoolEmAll include data center Simulation, Visualization and Decision tools (SVD Toolkit) and models of Data Center Efficiency Building Blocks (DEBBs). The resulting tools of CoolEmAll will permit planners and operators of data centers to carry out flexible and fast simulations to minimize the energy consumption on it and to reduce the associated greenhouse gas emissions. Several metrics have been proposed to assess the energy efficiency on data centers on the framework of CoolEmAll project. Unlike the common metrics on data center industry, the ones proposed in this project are focused not only on power consumption but also on dynamic heat-aware analysis. New metrics developed on CoolEmAll are (a) the Imbalance of Temperature of node, group of nodes and racks and (b) Rack Cooling Index adapted to a group of nodes. Node is defined as the smallest element of a data center to be modeled. This approach will permit to detect the cooling requirements and its source in order to implement strategies to reduce that energy demand. The paper describes the selected metrics and the results obtained on the CoolEmAll first prototype experimental tests. I.The evolution of our society towards the massive IT causes an exponential increase of needs on Cloud and HPC (High Performance Computing). Every year the IT (Information Technology) business grows and therefore new data centers are required. So far it is responsible for 2% of the total energy consumption worldwide and is still growing [1] [2]. The data centers are very intensive in energy consumption, mainly on power and cooling requirements. However, according to the survey [3], at least 50 % of respondent states that saving energy is a major priority for their organizations.The CoolEmAll project has the objective of improving energy efficiency of data centers providing a new design and real performance assessment tools [27]. Until nowadays, efficiency has been mainly driven by the minimization of the Power Usage Effectiveness (PUE) metric [4]. In most of the cases analysis are based on peak or average loads, static analysis and power consumption. This strategy faces many limits as it does not allow predictions of energy performance to improve the energy efficiency. Besides that, main problems on high energy consumption are related to cooling requirements. Only assessing the power consumption is difficult to detect where the issues associated to heat transfer has been originated. Some commercial approaches apply Computational Fluid Dynamics (CFD) simulations to analyze data centers' cooling efficiency but they often require advanced technical skills to be applied. Additionally, existing tools and metrics usually concentrate on peak or average loads do not permit to assess the data center efficiency in a transient status depending on several factors as level of load, application type and management policies.CoolEmAll tries to address these missing issues and facilitate a process of the det...

No abstract