This paper examines cache interference effects, shared resource contention, and the implications of shared resources on a multi-core processor. We study and characterize the performance degradation that occurs when processes compete for cache resources. This paper implements the concept of utilizing an aggressor or “red app” and a baseline or “blue app” testing as a means of testing worst-case scenarios and stressing the system while identifying interference channels, the pathways in the hardware that are confirmed to cause contention. Red apps are parallel applications executing on separate cores which degrade the performance of the blue app. First, we establish a baseline by running a baseline test app on one core while the other apps remain idle. This baseline app is called the blue app. Second, we run that app with the same configuration, but now adding an aggressor application or red app to a separate core. We determine the interference penalty and other significant effects by running multiple analysis tests. The same procedure can be repeated for different blue apps with combinations of different red apps.
Analysis of Cache Memory Interference on Multicore Systems Utilizing Shared Memory Aggressor Applications
29.09.2024
854369 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Online Contents | 1995
H.M.S. "dark aggressor" commissioned
Engineering Index Backfile | 1955
Proposal for a New Aggressor Aircraft
NTIS | 1988
|