Environment data processing for a data centre (2)

• المؤلف الرئيسي: Lua, Ivan Yangzhi
• مؤلفون آخرون: Tan Rui
• التنسيق: Final Year Project
• اللغة: English
• منشور في: Nanyang Technological University 2024
• الموضوعات: Computer and Information Science; Engineering; Data centre efficiency; Data analysis; Energy consumption behaviour; Predictive models; Machine learning models; Data-driven optimisation strategies
• الوصول للمادة أونلاين: https://hdl.handle.net/10356/174987
• المؤسسة: Nanyang Technological University

This final year project focuses on advancing the efficiency of data centre operations through the utilization of wireless sensor data. The project aims to explore the relationships between cooling system operation, server workloads, and the thermal environment within the data centre. The primary objective is to analyze the energy consumption behavior of multiple selected data centre components, developing models and algorithms to predict energy consumption patterns based on the identified correlations. We will leverage a methodology structured in two distinct, yet interconnected stages. Initially, the project aims to explore and understand the relationships between the cooling system operation, server workloads, and the thermal environment within the data centre. This stage involves comprehensive exploratory data analysis to establish pertinent correlations that can guide the optimization of cooling system control and server workload management. Building upon this fundamental understanding, the subsequent stage focuses on analyzing the energy consumption behavior of select data centre components. Utilizing the insights gathered from the first stage along with employing advanced techniques like the Filter Method and Wrapper Method for feature selection, we work towards developing robust predictive models and algorithms. Herein, the Embedded Method, which involves machine learning models such as Linear Regression, ElasticNet, Support Vector Regression, Random Forest, AdaBoost, CatBoost, XGBoost, and LightGBM, is used to capture complex relationships within the dataset and determine feature importance. The project anticipates delivering tangible outcomes in the form of enhanced data centre efficiency. By leveraging data-driven optimization strategies, the research aims to improve cooling system control, optimize server workloads, and provide a deeper understanding of energy consumption patterns. The anticipated results have practical implications for the broader field of data centre management, with potential energy savings and long-term sustainability as key drivers. The findings from this project contribute to the ongoing efforts to optimize data centre operations and address the challenges of energy consumption in the digital era.