X. Hu, T. Lei, Y.-C. Gong, and X.-H. Zhou (PRC)
RWCET estimation, dynamic scheduling, and power
Modern real-time applications often employ dynamic
scheduling to offer better system qualities, e.g., lower
power consumption, etc., while satisfying time constraints.
Their runtime decisions are mostly made based on the
tasks' remaining worst-case execution time (RWCET).
This paper first develop a novel analytic model to
formally express and solve the runtime RWCET
estimation problem, then introduce a series of experiential
methods to implement the formal solution by balancing
the estimation precisions against the runtime overheads.
By this means, it can be realized with a lower analytic
complexity to transform source codes to an adaptive
program, which can provide the scheduler at runtime with
its own RWCET estimation. Experimental results show
that the proposed method is effective and, expectably, can
be used as a foundation for other dynamic scheduling
techniques, e.g., adaptive power management, etc.