Poster Abstract: Enabling IoT Applications via Dynamic Cloud-Edge Resource Management

Emerging augmented reality IoT edge devices, such as smart glasses, show significant promise in use as assistive devices that involve human(s) in the loop. Owing to the often safety-critical nature of the assistive application, such as guiding a blind person, both the performance and dependability of the end-to-end system is critical. Among the many technical hurdles in realizing such applications, limitations in performance assurances from the cloud-hosted backend servers that support these applications, and the longer and often unpredictable end-to-end network latencies between the end user and the cloud can be detrimental to the response time requirements of such applications, which can compromise the safety of the end users. This poster presents ongoing work we are conducting that addresses these challenges by dynamically managing resources across the cloud-edge spectrum. Specifically, we describe the Dynamic Data Driven Cloud and Edge Systems (D
³
CES) approach, which uses measurement data collected from adaptively instrumenting the cloud and edge resources to learn and enhance models of the distributed resource pool, and in turn using these models in a feedback loop to make effective resource management decisions to deploy edge applications and deliver their QoS properties. The poster will also highlight a number of research directions we are pursuing in this project.