Cloud Management with Reinforcement Learning

PhD Qualifying Examination


Title: "Cloud Management with Reinforcement Learning"

by

Mr. Qizhen WENG


Abstract:

Cloud Computing, hiding the complexity of managing clusters from the 
developers, poses the challenges to cloud service providers to improve the 
quality of services while maintaining low operating cost. Through the 
years, many researchers have been dealing with these issues from different 
aspects, including network optimization, resource management, workload 
scheduling, etc. But the difficulty also arises with the growing scale of 
clusters and the heterogeneity of applications.

Inspired by the recent advances in artificial intelligence problems, the 
idea of applying reinforcement learning techniques to assisting the 
management of cloud systems is increasingly attractive. Instead of 
manually exploring vast configuration space in response to various 
workloads, cloud service providers can deploy learning agents to collect 
the data, interact with complex environments automatically, and improve 
the system's efficiency to the human expert-level or even beyond.

This survey serves a review of selected cloud management topics addressed 
by reinforcement learning approaches. We first give a brief introduction 
of fundamental concepts and advanced models of reinforcement learning. 
Then we present several series of applications in particular fields, i.e., 
network optimization, virtual machine configuration, dynamic power 
management, and cluster scheduling. After showing how researchers 
formulate system optimizations in different settings as reinforcement 
learning problems and design learning agents tackling real-world issues, 
we conclude the survey by discussing the common characteristics and 
challenges of these applications, motivating further research and 
industrial-oriented solutions.


Date:			Wednesday, 8 May 2019

Time:                  	10:00am - 12:00noon

Venue:                  Room 3494
                         Lifts 25/26

Committee Members:	Dr. Wei Wang (Supervisor)
 			Prof. Qian Zhang (Chairperson)
 			Dr. Kai Chen
 			Dr. Yangqiu Song


**** ALL are Welcome ****