IoT and Edge Computing

• Billions of distributed and network enabled smart objects (e.g. IoT, smart cars, mobile, etc) are connected to the internet, generating tremendous amount of data stream in edge environments. Traditional cloud computing paradigm is not sufficient to service all the data streams generated from the edge due to long latency in between the data provider, data analysis and service consumer. For video analytic service, for instance, large amount of video data generated from IoT cameras are transmitted to the cloud for analyzing the contents and service quality of the video data using various filtering algorithms. When service quality is critically affected due to change in environmental context of IoT data, quick detection and reconfiguration is required. However, if all analyzing operation is done at cloud, service quality will be degraded due to long latency.

Why do we need service-aware IoT virtualization in edge computing?

• Unlike general computing resources, IoT devices have various application-specific resources. They are still being used as common resources for multiple IoT application services. For example, a video frame captured at the same CCTV can be used for both tracking a suspecious human and detecting some objects like cars or bags. In order to virtualize application-specific resource of a resource-constraint device like IoT, lightweight virtualization is required. Partial application tasks on IoT devices / smart objects may detect if IoT resource degrades overall service quality and quickly react and change its configuration before it is analyzed at the cloud. Thus, we aim to build a novel service-aware IoT architecture which is capable of virtualizing IoT resources and reconfigure IoT resource’s configuration promptly upon service degradation to minimize application service quality deterioration time. This requires mapping between application service and IoT resource, a lightweight virtualization on IoT devices to share resources and reinforcement learning based resource reconfiguration algorithm to optimize service quality.

Core Research Topics to Realize Edge IoT Virtualiziation:

• An ontology-based knowledge for mapping various application service and IoT functionality
• Container-based lightweight virtualization scheme for micro-services on IoT
• IoT resource reconfiguration algorithm for situation-aware adaptation
• Optimazation of IoT resource reconfiguration using reinforcement learning
• Development of edge IoT cloud testbed