This research was aimed towards performance-power tradeoffs for real-time systems. Power constraints are becoming increasingly important as we have more complicated semiconductor devices, sophisticated mobile and space related applications. Scheduling has long been recognised as perhaps being the most important system management concept. We aim to develop innovative scheduling algorithms to achieve our power performance tradeoffs. The technology we target are processors that can operate at a finite number of discrete voltage levels.
There were three specific scheduling problems that were under investigation.
Minimizing energy for Hard Real-Time Periodic Task Graphs with Precedence Constraints.
Maximizing the accumulated reward for a set of Hard Real-Time IRIS tasks within a given energy budget.
Scheduling schemes for meeting Real-Time requirements of Soft Aperiodic tasks under a given energy budget and a given period for that budget.
To investigate these scheduling problems a in house simulator was under development. A collection of test inputs were also put together to create a consistent database for evaluating these and similar scheduling schemes.