Power Efficient Scheduling in Hard Real-Time Systems
on Multiprocessor Platforms
From satellites to personal computers, multiprocessors have taken over the computing industry. It is only a matter of time before the embedded
environment will have embraced the new trend. It won’t be without good reason. Multiprocessors can offer more performance while consuming
less power, provided the right task set.
The research focuses on identical multiprocessors, which support independent frequency scaling and dynamic voltage scaling. When the identical
multiprocessors are run at unique frequencies, the system resembles a uniform processing platform. This linear relationship among the processors’
computational capacities gives the scheduler the freedom to move tasks, without jeopardizing the task deadlines. The research focuses on selecting
the minimal processor frequencies that can complete the task set on time and then scheduling the tasks on these processors. The goal is to have an
online speed scaling and scheduling algorithm that can achieve an optimal
scheduler with the lowest possible power consumption.