Configuration Optimization of Fog Computing Platforms for Control Applications

Industry 4.0 requires the convergence of Operational and Information Technologies (OT & IT), which use different computation and communication technologies. Cloud Computing cannot be used for OT involving industrial applications since it cannot guarantee stringent non-functional requirements, e.g., dependability, trustworthiness and timeliness.

Instead, a new computing paradigm, called Fog Computing, is envisioned as an architectural means to realize the IT/OT convergence. A Fog Computing Platform (FCP) brings computing and deterministic communication closer to the network’s edge, where the machines are located in industrial applications. An FCP is implemented as a set of Fog Nodes (FNs) that integrate communication, computation, and storage resources.

Similar to previous research and ongoing standardization efforts, we assume that the communication between FNs is achieved via the IEEE 802.1 Time Sensitive Networking (TSN) standard. With the IT/OT convergence, applications of mixed-criticality will share the same FCP. At one extreme, we have the safety-critical real-time systems that control industrial process and have to be operational even in the case of failure.

The vision is to virtualize these as applications composed of tasks and messages running on an FCP. At the other extreme, we have non-critical Fog applications that do not have stringent timing and dependability requirements but are required to implement the novel functionalities of Industry 4.0. We assume that the platform uses partitioning to enforce the spatial and temporal isolation between applications with different criticalities.

Applications are modeled as tasks interacting via messages transmitted as flows on TSN. We consider several scheduling policies for tasks within a hierarchical scheduling model that can address the varied time-criticality requirements of applications. For example, the critical control applicaions are scheduled using static cyclic scheduling, and the resources of the Fog applications are allocated at runtime using best effort policies.

We propose several approaches to the design time FCP configuration optimization for mixed-criticality applications, such that the performance (in terms of Quality-ofControl) and timeliness of control applications are guaranteed, and the Quality-ofService of non-critical Fog applications is maximized.

In addition, we are interested in extensible configurations that support the addition of future new control applications and can successfully accommodate at runtime a large number of responsive Fog applications. At runtime, our approaches handle the migration and best–effort scheduling of Fog applications to the FNs that have resources for their execution.

Determining an FCP configuration means: deciding the partitions that provide temporal and spatial isolation among mixed-criticality applications, mapping the tasks to the cores of the multicore FNs, routing of flows on TSN, synthesizing the task schedule tables and the Gate Control Lists for switches that schedule the transmission of flows.

We have developed several algorithms that use heuristics, metaheuristics and Constraint Programming to solve these combinatorial optimization problems. The algorithms have been extensively evaluated on several test cases, including realistic test cases from the industry.