Towards Extensibility-Aware Scheduling of Industrial Applications on Fog Nodes

Fog computing has been identified as an enabler for many modern technologies like connected vehicles and the Industrial Internet of Things (IIoT). Such technologies are characterized by the integration of applications with different levels of criticality on shared platforms, which are referred to as mixed-criticality systems.

Mixed-criticality systems typically use static scheduling for critical tasks; however, static scheduling is not suitable for scenarios where fog nodes run dynamic noncritical applications that implement, e.g., maintenance checks and data analytics. To address this challenge, in this paper, we differentiate between critical tasks that are statically allocated (called “native”) and dynamic non-critical tasks that may migrate across fog nodes (called “temporary”).

We propose a static scheduling approach that maximizes the number of temporary tasks that can be added at runtime, without negatively impacting the already scheduled native tasks. This approach enables fog nodes to become more suitable for IIoT environments by configuring them with extensible schedules for the native tasks.

To evaluate our approach, we perform experiments considering several test cases, which show that given a number of native tasks, the generated extensible schedules enable the fog nodes to run a larger number of temporary tasks at the same time. Furthermore, the extensible schedules exhibit 7.8 % less missed deadlines (on averaae), compared to the non-extensible schedules.

To address this challenge, in this paper, we differentiate between critical tasks that are statically allocated (called “native”) and dynamic non-critical tasks that may migrate across fog nodes (called “temporary”). We propose a static scheduling approach that maximizes the number of temporary tasks that can be added at runtime, without negatively impacting the already scheduled native tasks.

This approach enables fog nodes to become more suitable for IIoT environments by configuring them with extensible schedules for the native tasks. To evaluate our approach, we perform experiments considering several test cases, which show that given a number of native tasks, the generated extensible schedules enable the fog nodes to run a larger number of temporary tasks at the same time.

Furthermore, the extensible schedules exhibit 7.8 % less missed deadlines (on averaae), compared to the non-extensible schedules.