A Survey on Serverless Machine Learning Systems

PhD Qualifying Examination


Title: "A Survey on Serverless Machine Learning Systems"

by

Mr. Yuheng ZHAO


Abstract:

The rapid evolution of cloud computing introduces serverless computing as a 
transformative paradigm, particularly in the deployment and scalability of 
machine learning systems. This paper surveys the integration of serverless 
computing within machine learning workflows, highlighting its advantages such 
as cost-efficiency, auto-scaling capabilities, and simplified operational 
management. We explore the dual phases of distributed machine learning: model 
training and model serving, emphasizing the unique challenges and opportunities 
presented by serverless architectures.

The paper reviews existing research on serverless model training systems, 
including general-purpose frameworks and specific training scenarios, as well 
as serverless model serving systems that address performance, 
cost-effectiveness, and management ease. We hope this survey can shed light on 
the potential of serverless computing in enhancing the efficiency and 
scalability of machine learning systems, while also identifying areas for 
future research and development.


Date:                   Monday, 26 August 2024

Time:                   2:00pm - 4:00pm

Venue:                  Room 3494
                        Lifts 25/26

Committee Members:      Dr. Wei Wang (Supervisor)
                        Dr. Shuai Wang (Chairperson)
                        Prof. Kai Chen
                        Prof. Song Guo