Adaptive QoS-Based Resource Management Framework for IoT/Edge Computing
Tom Springer, Erik J. Linstead · 2018
With the vast proliferation of Internet of Things, it will soon become unsustainable to manage the increasing number of connected devices in the cloud. An alternative is to perform a significant amount of the processing at the "edge" of the IoT cluster. Edge nodes can be placed in close proximity to the IoT devices to offload most of the processing between the devices and the cloud. However, with unpredictable processing requirements and varying workloads, care must be taken to adequately manage resources, so as not to overwhelm an edge node. For this reason, we present a new resource management framework for edge computing. Our method utilizes a Quality-of-Service (QoS) based approach to resource management that utilizes the concept of virtual clustering to dynamically allocate resources in a multiprocessor environment. In order to account for unpredictable computational loads, we incorporate feedback to adjust the resource levels based upon the QoS parameters. Fuzzy logic is used to combine various task properties such as, timeliness, resource utilization and criticality to make scheduling decisions. Simulations results indicate that this adaptive approach more effectively manages the system resources and provides for protections against task overruns during overload conditions as compared to traditional static allocation techniques.