Title: Mobile Device Clusters as Edge Compute Resources: Design, Deployment, and Role in the Computing Ecosystem

Karim Habak
School of Computer Science
College of Computing
Georgia Institute of Technology

Date: Thursday, February  1st, 2018
Time: 3 PM to 5 PM EST
Location: Clough Commons 150

Committee:
------------
Dr. Mostafa Ammar (Advisor, School of Computer Science, Georgia Tech)
Dr. Ellen W. Zegura (Co-Advisor, School of Computer Science, Georgia Tech)

Dr. Umakishore Ramachandran (School of Computer Science, Georgia Tech)

Dr. Ada Gavrilovska (School of Computer Science, Georgia Tech)

Dr. Khaled Harras (School of Computer Science, Carnegie Mellon University Qatar)

Summary:
-----------

Edge computing offers an alternative to centralized, in-the-cloud compute services. Among the potential advantages of edge-computing are lower latency that improves responsiveness, reduced wide-area network congestion, and possibly greater privacy by keeping data more local. However, widely deploying the needed edge-compute resources requires (1) provisioning the load introduced at various locations, (2) huge initial deployment cost and management expenses, and (3) continuous upgrades to keep up with the increase in demand. The availability of under-utilized mobile and personal computing devices at the edge provides a potential solution to these deployment challenges. In this thesis, we propose taking advantage of clusters of co-located mobile devices to offer an edge computing platform. Scenarios with co-located devices include, but are not limited to, passengers with mobile devices using public transit services, students in classrooms and groups of people sitting in a coffee shop. We propose, design, implement and evaluate the Femtocloud system which provides a dynamic, self-configuring and multi-device mobile cloud out of a cluster of mobile devices. Within the Femtocloud system, we develop a variety of adaptive mechanisms and algorithms to manage the workload on the edge-resources and effectively mask their churn. These mechanisms enabled building a reliable and efficient edge computing service on top of unreliable, voluntary resources. Our work also includes building a network measurement system that enables mobile devices to accurately and efficiently acquire knowledge of their network parameters while communicating with a variety of compute service providers. The measurements, acquired by our system, allow mobile devices to select the compute service provider that matches their demand and meet their target level of quality of experience. Our system also aggregate the measurements obtained by all the mobile devices and use them to reduce the measurement overhead and identify locations where edge resource deployment will be beneficial.