Title:  Function Delivery Network: A Dynamic Spatial-temporal Execution Orchestrator for Serverless Computing

Committee: 

Dr. Madisetti, Advisor   

Dr. Ji, Chair

Dr. Beyah

Abstract: The objective of the proposed research is to characterize and develop a dynamic shared-resource orchestration technique, called Function Delivery Network (FDN), that handles high-velocity, multi-region, resource-intensive events in serverless cloud computing. The FDN dynamically implements spatial and temporal allocation of events for functions in Function-as-a-Service (FaaS) to improve cross-region execution on the cloud in a scalable and efficient manner. Applications using FaaS typically focus on events that might share the same target resource (function) at a particular time. This thesis transforms a point-to-point service serverless model used by public cloud providers, where users and services interact one-to-one with a serverless function, to a many-to-one approach, while dynamically optimizing the resource usage of these events over time and space. This approach uses techniques such as batching, replication, pattern detection, prediction, and resource sharing and initial results are promising. Further, the FDN makes use of a novel Result-as-a-Service (RaaS) model to provide a library of functions customized to one or more business domains, such as real estate, banking, financial, and entertainment.