SMC-CROD Dataset

Sustainable Multi-Cloud Resource Optimization Dataset (SMC-CROD 2026), is a large-scale real-world-inspired operational telemetry dataset designed for sustainable cloud-computing research, carbon-aware orchestration, and federated multi-cloud intelligence. The dataset contains 485,716 operational records collected at 5-minute intervals from heterogeneous cloud infrastructures spanning AWS, Azure, Google Cloud, and private OpenStack deployments. The dataset represents realistic multi-cloud operational behavior under varying workload intensity, energy-consumption patterns, renewable-energy availability, thermal dynamics, and resource-allocation conditions. The dataset includes temporal context attributes such as hour, day-of-week, month, weekend indicators, and peak-hour conditions to model workload seasonality and operational variability across distributed cloud environments. Resource-utilization features include CPU utilization, GPU utilization, memory usage, disk I/O rate, network-bandwidth consumption, VM utilization rate, active server count, task queue length, container density, and resource idle time, enabling detailed modeling of cloud workload dynamics and infrastructure stress conditions. Energy and sustainability indicators are incorporated through power-consumption measurements, GPU power draw, cooling-energy usage, server temperature, Power Usage Effectiveness (PUE), electricity pricing, renewable-energy utilization, carbon intensity, green-energy availability, and regional carbon indices. These attributes support carbon-aware scheduling, sustainable cloud optimization, and energy-efficiency analysis under heterogeneous operational conditions. The dataset further contains cloud-network and orchestration features, including inter-region latency, cross-cloud bandwidth, workload migration cost, data-transfer rate, available compute nodes, SLA violation risk, QoS score, regional load balancing, energy-efficiency ratio, resource-allocation efficiency, and sustainable-operation indicators. These features capture the interaction between workload management, operational performance, and sustainability-aware orchestration decisions in distributed cloud infrastructures. Two target variables are provided. The first target, Carbon_Emission_CO2_kg, is formulated as a regression task for carbon-emission estimation. The second target, Resource_Allocation_Action, is designed as a multi-class classification task representing Local_Execution, Workload_Migration, and Delayed_Allocation strategies under varying carbon, workload, and resource conditions.