Title

RT-Xen: Real-Time Virtualization Based on Fixed-Priority Hierarchical Scheduling

Document Type

Technical Report

Department

Computer Science and Engineering

Publication Date

2010

DOI:

10.7936/K7DF6PG8

Technical Report Number

WUCSE-2010-38

Abstract

The ability to build systems of systems is vital to the real-time community as we endeavor to build increasingly sophisticated systems. Subsystems must be modular and isolated to ensure desired real-time performance when assembled by the system integrator. While virtualization has gained broad adoption and success for system integration in non-real-time domains, no existing virtual machine manager is designed to enforce real-time performance. On the other hand, while a rich body of promising theories on hierarchical real-time scheduling algorithms has been proposed in the literature to support composable real-time performance guarantees, they have not been incorporated and evaluated in real virtualized systems. We have developed RT-Xen, the first virtual machine manager designed to support real-time domains through hierarchical real-time scheduling. RT-Xen bridges the gap between virtualization and hierarchical scheduling by introducing a hierarchical real-time scheduling framework in Xen, the most widely used open-source virtual machine manager. Experimental evaluation revealed important insights on the feasibility and efficacy real-time virtualization: (1) RT-Xen can effective provide virtualization and scheduling to Linux guest OSes at 1~ms scheduling resolution (quantum) while incurring moderate overhead; (2) RT-Xen can provide real-time performance isolation against overrun domains; (3) RT-Xen significantly outperformed the standard credit-based Xen scheduler in term of real-time performance. RT-Xen represents a promising step toward real-time virtualization for system integration of real-time applications.

Comments

Full text not available.Permanent URL: http://dx.doi.org/10.7936/K7DF6PG8

This document is currently not available here.

Share

COinS