What is it about?
Fog computing brings computational power close to Internet of Things (IoT) sensors, often relying on low-cost, energy-efficient Single-Board Computers (SBCs) like Raspberry Pis. While Kubernetes has become the standard platform for orchestrating containerized applications, its default scheduler was designed for robust data centers. It relies on static resource allocations and simple heuristics, frequently overloading resource-constrained fog nodes and degrading service quality. To solve this, the authors introduce Chronos (Customizable Heuristic Resource Orchestrator for Node Optimization Scheduling), a lightweight, hardware-agnostic scheduling operator for Kubernetes and K3s. Instead of guessing node capacity, Chronos dynamically queries real-time telemetry and time-series performance data stored in InfluxDB. System administrators can customize placement policies using declarative YAML templates tailored to network-, storage-, or CPU-intensive workloads without writing new code. Deployed on a physical testbed of Raspberry Pi 4 and 5 nodes, Chronos significantly outperformed the default Kubernetes scheduler: * Reduced network latency by up to 23% for network-heavy services. * Cut disk write times by up to 42% (and read times by over 55%) on storage-intensive tasks. * Accelerated response times by up to 20% for CPU-bound computations. Crucially, Chronos achieves these gains with negligible overhead (less than 1% CPU and under 3.5% RAM in idle mode), proving that intelligent, adaptive container scheduling is practical even on low-cost hardware.
Featured Image
Why is it important?
Built specifically for resource-constrained Fog IoT: Overcomes the rigid assumptions of the standard Kubernetes scheduler on low-power Single-Board Computers without demanding costly hardware upgrades. Replaces static allocation with real-time telemetry: Uses actual live metrics (via Telegraf and InfluxDB) rather than static resource declarations, preventing noisy-neighbor bottlenecks and node saturation. Code-free customization via YAML: Enables administrators to implement application-specific scheduling heuristics (focusing on network, disk, or CPU) purely through flexible configuration templates. Validated on physical testbeds, not just simulations: Evaluated against realistic, concurrent workloads on physical Raspberry Pi clusters running K3s and Cilium CNI. Lightweight and production-ready: Operates as a native Kubernetes Operator with minimal memory and CPU footprint, automated CI/CD deployment scripts, and encrypted TLS communication.
Perspectives
Testing Chronos on a physical cluster of Raspberry Pi 4 and 5 devices proved that edge orchestration does not require sacrificing performance or stability. Looking ahead, our next milestone is evolving Chronos toward a fully autonomous, "zero-touch" scheduling engine that dynamically infers the optimal strategy without requiring manual administrator tuning.
Dra Carmen Carrión
Universidad de Castilla-La Mancha
Read the Original
This page is a summary of: Enhancing fog IoT container deployment: A customizable Kubernetes scheduler, Future Generation Computer Systems, March 2026, Elsevier,
DOI: 10.1016/j.future.2025.108195.
You can read the full text:
Contributors
The following have contributed to this page







