ABSTRACT

MAINTENANCE STRATEGIES FOR THE INFRASTRUCTURE OF CLOUD DATA CENTERS: A CASE STUDY

Over the past thirty years there has been a significant change in the concept of Data Center, which have evolved to accommodate the increasing needs of users, but the common denominator is now recognized as power and cooling. A typical mid-size Data Center has now a number of compute servers of the order of a couple of hundreds, and storage for several petabytes, but modern technology hardware use a lot of electrical energy and dissipate a lot of heat. In addition, Data Centers need to be properly maintained without interruption, to provide continuous services to web users. Server for Artificial Intelligence applications work with power-consuming GPU’s, thus leading to a distributed liquid-based cooling system, rather than an air-based one. Water pipes and electricity directly connected to IT hardware pose big maintenance problems.
Proper maintenance strategies have to be formulated for an efficient Data Center with a sustainable Power Usage Effectiveness parameter, that is reducing the energy devoted to cooling only, and this paper describes a case-study  at the university Federico II, in southern Italy, where a 1 Mwatt Data center is maintained operational and efficient over two decades now, with a complete, distributed monitoring system and a particular attention to the power and cooling system.