Supplement of

“Thermal-Aware Management for

Embedded Real-time Systems” 64.6% by reducing core frequency to 0.8GHz in
order to meet the thermal constraint under the
Paper #92
thermal constraint of 55°C. RT-TRM is shown to
effectively control the processor temperature close
Job-level temperature impact under RT-TRM
to the thermal constraint and maximize the
We have included a new figure that shows task’s
resource utilization.
periodic execution and its effect on the processor
temperature. In particular, we have looked closer
the results shown in Fig. 7 in a shorter time interval
(0,40] and have presented the execution behavior
of the hottest task (bit manipulation) and its
corresponding processor temperature under RT-
TRM. We note that task-level power dissipation,
activity factor, and execution time with the
min/max period for each task are shown in Figure
2, Table 1, and Table 3, respectively. In the figure,
we observe that the processor temperature is
increased whenever the bit manipulation task is
executed. When the ambient temperature is 35°C,
a job of the bit manipulation task is invoked every
12 seconds, while it is invoked every 6 seconds
when the ambient temperature is 25°C. When the
ambient temperature is 30°C, it is invoked every 8 Workloads with different power dissipation
seconds. We have conducted additional experiments for
different power dissipation workloads under the
thermal constraint of 60 ℃. For the low power
dissipation workload, RT-TRM achieves a higher
task-rate of 88.9% with the maximum core
frequency. For the high power dissipation
workload, RT-RTM dynamically adjusts task-rate in
order to meet the thermal constraint and achieves
a task rate of 79.3%.

Different thermal constraints

We have conducted additional experiments for
other thermal constraints of 55 and 65°C. Under
the thermal constraint of 65°C, RT-TRM achieves a
higher task-rate of 91% without reducing core
frequency, while it achieves a lower task-rate of