Optimizing Placement Policies for DAG Function-as-a-Service Workloads in Edge Computing Environments
Author(s)
Sinha, Kartik
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
We explore scheduling policies as part of our proposed system FaaSEdge to enhance Edge computing by enabling Edge-native application development through the function-as-a-service (FaaS) paradigm. Addressing challenges in Edge development, FaaSEdge introduces front-end abstractions for seamless integration with Cloud Native development and streamlines back-end management operations for improved multi-tenancy. We explore the effects of function placement policies on bandwidth consumption. The study evaluates three function placement policies on a San Francisco taxi trace with edge nodes on city cell towers. Our results reveal significant bandwidth consumption disparities among policies, emphasizing the limitations of static placements. We show dynamic placement policies demonstrate a 4x reduction in net bandwidth consumption per-link, showcasing FaaSEdge's potential to optimize Edge computing for real-time applications.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis