CPU Shielding:
Investigating Real-Time Guarantees Via Resource Partitioning

 

Scott Tillman
CSC 714 Real-Time Computing Systems
Spring 2009

 

Objectives

This project seeks to investigate the feasibility and limitations of using CPU shielding to allow hard real-time operation in commercial, off-the-shelf (COTS) systems. This will be done by bounding and verifying worst case interrupt response times (CPU contention), worst case bus reaction times (bus contention), and worst case slowdown associated with additional cache misses (cache contention). There are existing documents which discuss various models of these delays. The goal is to verify the models/predictions and evaluate the predictability that can be achieved using this co-hosting method.

Documents and Timeline

March 16th   Project Proposal
March 22nd Replicate system setup from [Brosky]. Evaluate literature predictions of interference from known sources (PCI, Interrupt and Cache induced)
March 29th Design/replicate experiments to test latency from known sources
April 2nd Initial Project Status Report
April 5th Gather and evaluate initial test results. Identify and categorize unknown latency factors.
April 19th Demonstrate mixed (real-time and non-real-time) mode operation at predicted highest frequency
April 20st Final Project Summary Slides
April 21st Final Project Status Report

Project Files

Can be found here.

Reference Literature & Links