New IoT-Fog-Cloud Scheduling Dataset Released
The labeled data enables developers to train machine learning models for resource allocation in distributed computing networks.
Updated on Oct. 27, 2026 in Data Centers
Live Poll
Do you trust automated systems to manage resource scheduling for critical technology infrastructure?
Researchers have released a labeled dataset designed for training machine learning models to manage resource scheduling in IoT-fog-cloud architectures. The dataset, generated through a genetic algorithm, provides a foundation for developing automated scheduling systems in distributed computing environments.
Why it matters
Efficient resource allocation is critical to managing modern distributed computing, but ML-based schedulers have historically lacked the labeled training data required for effective deployment. This release provides a benchmark for testing new scheduling algorithms across heterogeneous compute nodes.
The FogMLS dataset includes 404,919 task-resource assignments within its high-load configuration, featuring 200 task positions per scheduling slot. A Random Forest model trained on this data achieved a consistent accuracy range of 76.14% to 76.35% across three different random seeds.
The details
The dataset maps task characteristics—including instruction count, data size, and node capacity—to resource assignments within a three-tier IoT-fog-cloud simulation. It was compiled using a genetic algorithm, an optimization method that mimics natural selection to find high-performing solutions, paired with greedy initialization, a strategy that makes locally optimal choices at each stage of a process. These methods provide a standardized set of labels for training future predictive schedulers.
Timeline
- 2026-10-27
The research paper and FogMLS dataset were officially released.
The Tech Race
This release addresses the data scarcity hindering the development of autonomous schedulers in edge-to-cloud environments. It creates a standardized benchmark against which future reinforcement learning or heuristic scheduling agents can be measured.
Developers and researchers building distributed computing management tools can now utilize this dataset to train and validate their scheduling models. It serves as an immediate, ready-to-use resource for benchmarking new algorithms without needing to build custom simulation environments from scratch.
The takeaway
The availability of this dataset removes a common barrier to testing machine learning-based resource schedulers in distributed environments. Researchers should monitor whether future scheduling models exceed the 76.35% accuracy ceiling established by the Random Forest baseline in this initial release.
Further reading
Learn more about the infrastructure trends impacting current Data Centers.
Source note: This article includes information reported by Nature.
Live Poll
Do you trust automated systems to manage resource scheduling for critical technology infrastructure?
