Environment data processing for a data centre

• Main Author: Joshi, Anirudh
• Other Authors: Tan Rui
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/74038
• Institution: Nanyang Technological University

Data centres have become essential with the increasing popularity of cloud computing. Server overheating is a major problem faced by these computing infrastructures. Overcooling server rooms is a common practice to prevent overheating. However, this strategy results in excessive power consumption. This project aims to build an analytical model which can accurately predict server temperature. With an earlier indication of potential server overheating, preventive strategies can be adopted. The objective is to reduce the reliance on overcooling and improve temperature control systems to make them more energy efficient. The environment data used in this project was collected in a data centre using various wireless sensors (Chen, et al., 2014). Linear regression and random forest regression techniques were applied to understand the relationship between central processing unit (CPU) utilisation, CPU power consumption and data centre environment temperature. These analytical models were used to predict temperatures at varying points of time in the future (time horizons). The models were fairly accurate in predicting the environment temperature. Random forest regression performed better with an average root mean squared error (RMSE) of 0.369 at a time horizon of 10 minutes, compared to linear regression’s 0.412. Although a decent indication of the risk of overheating, the temperature predictions are limited by the restricted number of independent variables analysed. The thermal environment is affected by various factors, more of which should be included in the analysis to better understand the interrelationships and increase the prediction accuracy.