Context-Aware Kubernetes Scheduler for Edge-native Applications on 5G

Abstract

This paper is an extension of work originally presented in SoftCOM 2019 [1]. The novelty of this work reside in its focused improvement of our scheduling algorithm towards its usage on a real 5G infrastructure. Industrial IoT applications are often designed to run in a distributed way on the devices and controller computers with strict service requirements for the nodes and the links between them. 5G, especially in concomitance with Edge Computing, will provide the desired level of connectivity for these setups and it will permit to host application run-time components in edge clouds. However, allocation of the edge cloud resources for Industrial IoT (IIoT) applications, is still commonly solved by rudimentary scheduling techniques (i.e. simple strategies based on CPU usage and device readiness, employing very few dynamic information). Orchestrators inherited from the cloud computing, like Kubernetes, are not satisfying to the requirements of the aforementioned applications and are not optimized for the diversity of devices which are often also limited in capacity. This design is especially slow in reacting to the environmental changes. In such circumstances, in order to provide a proper solution using these tools, we propose to take the physical, operational and network parameters (thus the full context of the IIoT application) into consideration, along with the software states and orchestrate the applications dynamically.

Container Placement and Migration in Edge Computing: Concept and Scheduling Models

“…The Kubernetes scheduler taint/toleration and predicates/priorities mechanisms are also used to restrict which pod is placed on an edge node [73]. Chima et al [77] give a good account of the Kubernetes scheduler backend functions.…”

“…They formulate the requirements as a multiobjective optimization problem, solved as an integer linear programming problem. A model proposed by Chima et al [77] aims to take the physical, operational, network and software states into consideration. The authors apply a hybrid scheduling model combining both the queuing model and the concurrent model.…”

“…Kubernetes allows custom scheduling logic and decision variables to be added to the default model in a variety of ways [9], but this can create additional computation overhead for the scheduler [77]. A custom scheduler can be built and implemented outside of Kubernetes control and made to interface with Kubernetes through an external module [77]. The external custom scheduler runs side-by-side the default Kubernetes scheduler and either of the two schedulers is invoked depending on the requirement.…”

Voilà: Tail-Latency-Aware Fog Application Replicas Autoscaler

“…Zheng et al propose to vary the number of pods according to the load, but does not address the question of pod placement nor efficient routing between the end users and their closest pod [27]. Several other works propose location-aware pod placement mechanisms for Kubernetes [4], [8], [14], [24]. However, they consider the number of replicas as a constant and do not aim to vary it to accommodate non-stationary workloads.…”

“…Voilà integrates seamlessly with Kubernetes, the de-facto standard container orchestration framework in clusters and data centers [12]. Kubernetes is also a promising basis for designing future-generation fog computing platforms [4], [7], [14], [24], [27]. Voilà continuously monitors the request workload produced by all potential traffic sources in the system, and uses efficient algorithms to determine the number and location of replicas that are necessary to maintain the application's QoS within its expected bounds despite potentially large variations in the request workload characteristics.…”