Context-Aware K8S Scheduler for Real Time Distributed 5G Edge Computing Applications

The problem of service availability assurance and optimal performance in 5G edge computing is solved by means of static or rudimentary dynamic scheduling techniques. In the cases where Kubernetes orchestration is used, the default scheduler manages the allocation of services with methods designed to suit general cloud application requirements, not specific to the requirements of edge computing (i.e. evaluating basic usage devices parameters). This design has its shortcomings, especially due to its slow reactions to the environmental changes. In such circumstances, it would be preferred to take the physical, operational and network parameters into consideration along with the software states and orchestrate the applications dynamically. We propose an improved design that integrates real-time information about the edge devices in the scheduler decision algorithm enabling advanced orchestration and management along with improved fault tolerance capabilities of the network edge. We validate the utility of the design by comparing it to the default Kubernetes scheduler.