HRRMLQ: Container scheduling algorithm on edge nodes cluster

With the evolution of cloud computing, the dominant virtualization technology for data centers has shifted from virtual machine to container. Despite its success, Cloud Computing falls short of several Internet of Things requirements (IoT). Edge Computing appears to be a complement to the Cloud, filling gaps in the IoT scene. By 2025, it is expected that 75 percent of the generated data will be processed in locations other than cloud servers or datacenters. As a result, the number of edge computing nodes is expected to grow rapidly. Among others, Kubernetes was created with the goal of automatically deploying and managing cloud applications created using container runtime techniques such as Docker. Although Kubernetes is designed for cloud computing environments, it does not have adequate system support for edge computing. To use Kubernetes in edge computing environments, edge computing nodes must be managed in a more efficient manner. The Hybrid Round-Robin multi-level queue scheduling algorithm (HRRMLQ) is proposed and implemented in this paper to schedule containers to a single board computer (SBC) cluster based on their priorities. HRRMLQ uses a hybridization of round robin and multilevel queue scheduling algorithms to categorize IoT apps as high or low priority. HRRMLQ classifies IoT apps as either high or low priority. To alleviate the starvation problem that traditional multilevel queues suffer from, HRRMLQ assigned 4 quantum to the high priority queue and 2 quantum to the low priority queue. Experiments show that HRRMLQ can schedule high priority pods first while also reducing the starvation problem that multilevel queues may experience.