Thermal Feasibility Analysis of Periodic Real-Time Tasks with Jitter
Date
2014-08-24Author
Bhimsaria, Devesh
Advisor(s)
Ramanathan, Parameswaran
Metadata
Show full item recordAbstract
It is well-known that spatial and temporal variations in power density within a processor core creates thermal hot spots where the temperatures can exceed 100 C. Such temperatures can degrade device reliability and adversely affect processor performance. This problem is expected to become even more acute in the coming years because average power densities are expected to double in the next decade. As a result, estimating the worst-case peak temperature for a given processor workload is an important problem.
This report presents an algorithm to estimate the worst-case peak temperature for a given real-time workload in a multi-core system. The challenge is to find the worst-case arrival pattern for the given real-time workload that results in the maximum peak temperature when the tasks are executed using a work-conserving scheduler. The proposed algorithm improves a solution from literature by relaxing some of its restrictive assumptions. In the cases where the assumptions are satisfied, the estimates from the proposed algorithm and the scheme from literature are the same. In all other cases, the estimates from the proposed algorithm are much less conservative and hence, more accurately estimate the achievable worst-case temperature.
Subject
real-time systems
multi-core systems
multiple task set
worst case peak temperature
thermal analysis
Permanent Link
http://digital.library.wisc.edu/1793/70080Part of
Licensed under: