Resource Utilisation and Energy Consumption Telemetry for a Weather Forecasting Scenario in the ENACT Continuum

The dataset is described in this paper: https://doi.org/10.1016/j.dib.2026.112734 If you are using this dataset, please cite: Kapetanidou, Ioanna Angeliki, Thanasis Kotsiopoulos, Giorgos Thanasoulis, Paschalis Bizopoulos, Athanasios Liatifis, Manolis Skoularikis, Alexandros Nizamis, Panagiotis Sarigiannidis, and Konstantinos Votis. "A Telemetry Dataset on Resource Utilisation and Power Consumption in the Edge-Cloud Continuum." Data in Brief (2026): 112734. Dataset Overview This dataset provides time-series telemetry measurements of resource utilisation and energy consumption across an edge–cloud computing environment. The data were generated from a distributed weather forecasting application deployed on a Kubernetes cluster wherein other services are also running. Therefore, telemetry data were obtained under real-world operational conditions rather than in a controlled, isolated testbed. Data are collected for two computing nodes: a cloud-based VM (vm-node) and a Raspeberry Pi 4 edge device (rpi4), to reflect different computing capabilities five K8s workloads: two weather data collection pods, each corresponding to a data source from a different city (namely, Thessaloniki and Berlin), deployed at the edge two AI-based weather forecasting applications, one deployed at the cloud and one at the edge node a MinIO storage pool deployed on the cloud node for long-term storage of data The data include time-series measurements of node-level metrics, CPU, memory, and disk utilisation, network throughput, and energy consumption for the infrastructure nodes, as well as pod-level metrics, i.e., CPU and memory utilisation, and energy consumption. Telemetry data was collected during two distinct phases: During the first period, from 24 January 2026 to 26 January 2026, all pods were deployed and running continuously for three consecutive days. During the second period, from 31 January 2026 to 2 February 2026, all application pods except for the MinIO storage pod were scaled down to zero replicas. This allows for capturing the system baseline behavior and assessing the impact of workload activity. The dataset consists of four CSV files: 1. node_telemetry_pods_on.csv contains time-series measurements collected from infrastructure nodes for the first period (24/01/2026-26/01/2026), when pods were up2. pod_telemetry_pods_on.csv contains time-series measurements collected by monitoring the deployed pods for the first period (24/01/2026-26/01/2026), when pods were up3. node_telemetry_pods_off.csv contains time-series measurements collected from infrastructure nodes for the second period (31/01/2026-02/02/2026), when pods were down4. pod_telemetry_pods_off.csv contains time-series measurements collected by monitoring the deployed pods for the second period (31/01/2026-02/02/2026), when pods were down Each row corresponds to a single timestamped observation for a specific node or pod. In node telemetry files, the columns are explained in the table below: Column name Description Unit node_name Node identifier vm-node or rpi timestamp Measurement timestamp DD-MM-YYYY HH:MM:SS CPU (%) Node CPU utilization Percent (%) MEM (%) Node memory utilization Percent (%) fs (%) Filesystem utilization of the node Percent (%) Energy (watts) Power consumption of the node Watts rx (B/sec) Incoming network traffic Bytes per second tx (B/sec) Outgoing network traffic Bytes per second In pod telemetry files, the columns are explained in the table below: Column name Description Unit pod_name Pod name weatherforecaster-berlin-edge or weathercollector-berlin or weathercollect-thess or weatherforecaster-berlin-vm or forecaster-tenant-pool timestamp Measurement timestamp DD-MM-YYYY HH:MM:SS CPU (%) Pod CPU usage relative to node CPU capacity Percent (%) MEM (B) Pod memory usage out of total node memory Bytes Energy (watts) Energy consumption attributed to the pod Watts