A Comprehensive Monitoring, Visualization, and Management System for Green Data Centers

Abstract

Maintaining service reliability, achieving sustainability, and ensuring energy efficiency are crucial for green high-performance computing systems. Balancing these factors is a key challenge for modern green data centers. In this research, we propose a monitoring, visualization, and management system for green data centers (MVMS-GDC). Our comprehensive automated platform includes “monitoring system” and “rules and policy management” modules. The “monitoring system” gathers and visualizes time series data from all resources of a green data center, tracking essential metrics and measurements. It audits green energy, microgrid, climate conditions, workloads, hardware, CPU and memory usage, cluster component health, computing node activities, and network health and quality metrics. The “rules and policy management” module defines and enforces policies to balance resources, ensuring a reliable, sustainable, scalable, and efficient computing environment. We implemented, tested, and evaluated the MVMS-GDC system using green energy at the Zephyr data center located at the GLEAMM site. Our results demonstrate at least a 4.9% improvement in performance, at least a 4% increase in energy efficiency, and a reduction of at least 4% in job losses. The MVMS-GDC system also enhances scalability by employing a policy machine for each compute node, which automates power state control (on, off, or hibernation) based on monitoring observations. This automated approach ensures efficient and dynamic scaling, making MVMS-GDC suitable for large and highly distributed data centers. Overall, MVMS-GDC provides a robust solution for balancing energy availability and computational needs, optimizing performance, and maintaining energy efficiency in green data centers.