Feedback-Based Resource Management for Embedded Multicore Platforms

Researchers: Karl-Erik Årzén, Martina Maggio, Enrico Bini, Georgios Chasparis

Funding: VR

Duration: 2012 - 2015

This project is aimed at advancing the state of the art in dynamic resource management
for embedded multicore computing platforms by applying control theory.
Efficient resources usage is becoming one of the most important design criteria
for all types of computer systems from large data centers over laptops, cellular
smart phones, and embedded computing devices down to sensor network
nodes. The overall goal of the resource management can be to minimize power
consumption or generated heat, or to have better means for differentiating applications
against each other, e.g., ensure that a safety-critical application is guaranteed
sufficient resources also in the presence of less important applications.

In embedded systems there are many resources that need to be managed, e.g.,
memory, buses, and power. The most important resource, however, is the CPU.
By controlling how much and where different applications may execute it is also
possible to control the power consumption and the heat generation. Hence, the
focus of this project is control of CPU resources.