Zusammenfassung
Predictable execution enforcement for applications
with highly and arbitrarily fluctuating resource usage requires
runtime resource management. Correct runtime predictions
regarding resource usage of individual components allows
making proper resource reservations, enabling a better
resource management of the component-based applications.
This work presents a state-based resource usage model for a
component, in which states represent CPU utilization
intervals. This resource model is intended for a resource-
aware component framework where it will be used to determine
the quality of resource reservation. For this purpose, the model
offers two metrics: failure rate, which measures the fraction
of the reservation periods for which the reserved budget was
insufficient, and resource waste, which measures unused
budget.To illustrate the model, we apply it to a family of reser-
vation prediction strategies and validate the outcome by means
of a series of experiments in which we measure the resource
utilization of two video components. The latter requires a
method for monitoring resource states which is also presented,
analyzed and validated in this paper.
Nutzer