Simulator for modeling, analysis, and visualizations of thermal status in data centers

Data centers (DC) are richly instrumented systems consists of highly coupled elements that store and process a large amount of data. To perform large computation and storage, a DC is equipped with more than thousands of servers or even more. Due to a large number of these computational devices put in use at DC, produces a large amount of heat. Therefore the cost to maintain the thermal balance in a DC has increased significantly and has become almost equal to the cost of operating these systems. The main problem in heat management is 'Hot spots creation' which can cause hardware inefficiencies, and operational disruptions and in turn have a negative impact on overall functionality. To address these issues this paper aims at decreasing energy consumption of DC by allowing administrators, designers, and planners to model, visualize and analyze the thermal status of various configurations and solutions. A major difficulty in the thermal analysis of DC is the lack of simulation tool, where the impact of design (layout) and workload on thermal status can be tested. Therefore there is a need for a simulator that approaches the problem from an end-user perspective and takes into account all the factors that are critical to analyzing thermal balance in DC. A simulator is developed that takes DC computational devices as input and provide models that allow designers to analyze and visualize DC thermal conditions. The simulator allows the user to apply different job allocation strategies and can analyze thermal status for each; that help them choose the best strategy for job allocation. It enables DC administrator to organize servers and racks before their actual implementation. Moreover, servers can be relocated to analyze and maintain thermal balance. The user will also able to predict thermal condition after a specific time period. The simulator enables a DC user/administrator to maintain thermal balance in DC by detecting the root cause of hot spots under different workload; which will help them make better workload balancing decisions. (C) 2018 Elsevier Inc. All rights reserved.