In many real-time applications involving human-computer interactions, the quality of service also depends on the way performance is changed during workload variations. When human factors affect performance measurements, smooth rate transitions are always preferred with respect to abrupt parameter changes.

In this paper, we propose a new methodology for automatically achieving smooth rate adaptation of a periodic task set during workload variations due to abrupt environmental or systems changes. Load balancing is performed using an elastic task model, according to which tasks utilizations are treated as damped springs with given elastic and damping co-efficients. The model has been implemented on top a real-time kernel and experimental results are reported to show the effectiveness of the proposed approach.