Master Thesis Defense: Khodayar Jeirroodi
Speaker: Khodayar Jeirroodi
Supervisor: Dr. B. Jaumard
Examining Committee: Drs. T.-H. Chen, D. Goswami, W. Shang (Chair)
Title: Efficient Heuristics for Virtual Machine Migration in Data Centers
Date: Thursday, July 11, 2019
Place: EV 3.309
Live migration of Virtual Machines is one of the essential virtualization technologies, which enables the consolidation and load balancing in cloud data centers without interrupting the services.
Main goals for optimizing a single Virtual Machine live migration is to minimize migration time, transferred data and downtime. Planning multiple live migrations in a data center has an essential impact on feasibility of consolidation and quality of services during migrations, however, optimizing parallel VM migrations has been studies less. Minimizing makespan (total migration time) while reducing energy and service quality degradation caused by using datacenter resources for migrations, are the main objectives of the problem. One of the issues in planning multiple live migrations is to detect and consider migrations order dependency constraints and possible deadlocks caused by lack of enough free resources in servers during the process.
In the literature, exact mathematical models are not scalable and heuristics are not optimal and they don't consider the quality of service and energy efficiency of migration process when resources are restricted.
In this work we propose a heuristic algorithm for scheduling the migration of virtual machines in a data center to minimize makespan and solve the conflicts (deadlocks) caused by limitation of resources with minimum cost and quality degradation.