Performability Modeling: Back to the Future?
J. F. Meyer
Abstract
After more than
30 years of work concerning its theory, techniques, tools, and applications,
performability modeling is well understood by the many
people who have been responsible for its development.
During this period, other concepts
have emerged which likewise aim to express how well a system performs (serves
it users) under realistic operating conditions that include occurrences of both internal
and external faults.
The most prevalent of these are various concepts and measures of quality of service
(QoS) and experience (QoE) which are "performability-like" in that they refer to aspects of
both system performance (in the strict sense) and dependability.
To make a more precise comparison with performability measures,
it would be helpful to observe some basic properties of the latter
which follow from the original modeling framework. In turn,
differences revealed in this comparison could point
to certain measure-formulation and model-solution problems
that deserve further attention. Accordingly, the title of this talk is
asking whether it's time to go back to
first principles and, after doing so, suggest what needs to be done to
facilitate future work on model-based QoS/QoE evaluation. Presuming a
"yes" answer to this question, both avenues are followed in the talk.
PDF File: PMCCS-8.pdf.zip