Research Area: Cloud Computing
Cloud and Edge Data Centers have become the backbone infrastructures of daily social and economical activities. Live migration is the cornerstone of the dynamic resource management policies for various objectives, such as application performance, networking cost, load balancing, consolidation, energy saving, user mobility, no-downtime maintenance, and disaster recovery. Live migration of VMs and containers provides a universal state-transfer standard to implement these objectives.
Therefore, it is critical to manage the live migration in both computing and networking resources to guarantee the QoS, improve migration performance, and minimize migration costs and overheads. Many works have focused on the live migration mechanisms and optimization to improve the performance of individual migration. However, existing migration models in resource management neglect the resource competitions and dependencies among multiple migrations. Furthermore, performing the generated multiple live migrations in arbitrary orders can lead to service degradation. Therefore, efficient migration generation and scheduling are essential to reduce the impact of live migration overheads and improve migration performance. In addition, to prevent Quality of Service (QoS) degradations and Service Level Agreement (SLA) violations, it is necessary to respect the deadline of migration requests with various priorities and urgencies.
In this thesis, we focus on network-aware multiple migration management based on Software-Defined Networking (SDN). By separating the control plane and forwarding plane, SDN provides centralized topology discovery and networking management which enables the capability of managing resource contentions in finer granularity.
Name of the Researcher: TianZhang He
Name of the Supervisor(s): Rajkumar Buyya
Year of Completion: 2021
University: The University Of Melbourne
Thesis Link: Home Page Url