[Defense] Resource Partitioning in Real-Time Virtualized Systems: Optimization, Implementation, and Application.
Friday, March 25, 2022
4:00 pm - 6:00 pm
In
Partial
Fulfillment
of
the
Requirements
for
the
Degree
of
Doctor
of
Philosophy
Guangli
Dai
will
defend
his
dissertation
Resource
Partitioning
in
Real-Time
Virtualized
Systems:
Optimization,
Implementation,
and
Application
Abstract
The application of the virtualization technique in critical scenarios with real-time performance requirements is increasingly popular. For the Virtual Machine (VM) scheduling in real-time virtualized platforms, a Hierarchical Real-Time Scheduling (HiRTS) framework is proposed. The framework divides the scheduling problem into task-level scheduling, which schedules tasks inside each VM, and resource-level scheduling, which schedules VMs on the physical resources. Accordingly, the Regularity-based Resource Partitioning (RRP) model defines a resource interface specification that communicates the task level and the resource level. To make the RRP model more practical to various scenarios, this dissertation solves the critical open problems left in the RRP model from three aspects: optimization, implementation, and application. Specifically, as task-level optimization, we propose more task-level schedulability tests applicable for more different types of tasks. We also conduct resource-level optimization of the RRP model by introducing fault-tolerant RRP and more efficient resource-level resource partitioning algorithms with proven performance bound. With the optimization that increases the efficiency and the viability of the RRP model for industrial usage, we implement the RRP model on hard-real-time hypervisor XtratuM and soft-real-time hypervisor Xen. Numerous experiments and case studies are conducted to validate the efficacy of the implemented RRP model and to discuss important parameter choices. Last but not least, we discuss the potential application of the implemented RRP model in lowering the cost of irregular High-Performance Computing (HPC) applications and in building the Cyber-Physical Space (CPS) for traffic routing to reduce road congestion.
Friday,
March
25,
2022
4PM
-
6PM
CT
Online
via
Dr. Albert M. K. Cheng, dissertation advisor
Faculty, students and the general public are invited.
