Full Scaling Automation for Sustainable Development of Green Data Centers

TL;DR

The paper tackles the rising energy footprint of data centers by addressing the need for accurate workload forecasting and stable, uncertainty-aware autoscaling. It introduces Full Scaling Automation (FSA), combining a representation-enhanced multi-scale time series forecasting model with a task-conditioned Bayesian neural regression-based scaling policy that yields upper and lower bounds for resource allocation. Through extensive real-world experiments, FSA achieves superior forecast accuracy, precise CPU-utilization estimation, and energy-efficient Pod scaling compared to state-of-the-art baselines, and it reports substantial carbon-emission reductions in industrial deployments. The work demonstrates a practical pathway toward greener data centers by integrating advanced representation learning, probabilistic decision-making, and scalable deployment in production environments.

Abstract

The rapid rise in cloud computing has resulted in an alarming increase in data centers' carbon emissions, which now accounts for >3% of global greenhouse gas emissions, necessitating immediate steps to combat their mounting strain on the global climate. An important focus of this effort is to improve resource utilization in order to save electricity usage. Our proposed Full Scaling Automation (FSA) mechanism is an effective method of dynamically adapting resources to accommodate changing workloads in large-scale cloud computing clusters, enabling the clusters in data centers to maintain their desired CPU utilization target and thus improve energy efficiency. FSA harnesses the power of deep representation learning to accurately predict the future workload of each service and automatically stabilize the corresponding target CPU usage level, unlike the previous autoscaling methods, such as Autopilot or FIRM, that need to adjust computing resources with statistical models and expert knowledge. Our approach achieves significant performance improvement compared to the existing work in real-world datasets. We also deployed FSA on large-scale cloud computing clusters in industrial data centers, and according to the certification of the China Environmental United Certification Center (CEC), a reduction of 947 tons of carbon dioxide, equivalent to a saving of 1538,000 kWh of electricity, was achieved during the Double 11 shopping festival of 2022, marking a critical step for our company's strategic goal towards carbon neutrality by 2030.

Figures (4)

• Figure 1: Model Architecture. Red dashed lines represent the workload forecast module, including Multi-scale Time Series Representation and Representation-enhanced Deep Autoregressive Model. Blue dashed lines highlight the scaling decision module via task-conditioned Bayesian neural regression. According to service tolerance for response time(rt), the optimal value between upper and lower bounds is obtained.
• Figure 2: Visualization of TS representation of two weeks (every 10min). Changes in daily and weekly periods are clearly indicated.
• Figure 3: The performance(lower is better) of different autoscaling approaches. The vertical axis represents "Autoscaling method", and the horizontal axis represents "Relative resource consumption-RRC%".
• Figure 4: The Comparison of Electric Energy Consumptions. The red line represents the electricity consumption of baseline, and blue line is the electricity cost of FSA. The green histogram is the carbon emission reduction using FSA over the past four years.