Challenges and Experiences in Transitioning Management Instrumentation from Command-Oriented to Model-Driven - Advances in Network-Embedded Management and Applications - page 3

and differences between the implicit and imposed models.
A model-driven system requires an underlying implementation in order to
access instrumentation data and services. This cannot be easily leveraged
from the command-oriented implementation due to its parallel nature.
Retrofitting model-driven instrumentation on a legacy application constructed
with a hard-coded command-oriented instrumentation is difficult. Defining an
accurate model based on the actual implementation is the most pressing
problem. Designers are faced with the choice of whether to use existing
implementations preserving their inconsistencies and redundancies or to
address these problems by creating a new streamlined model-driven
implementation from their domain knowledge and experience.
In this paper the understanding of how management instrumentation system
designs are impacted by this transition is discussed. Challenges and
experiences will be examined through the prism of a real-world development
effort of transitioning a command-oriented management instrumentation
system to a model-driven management instrumentation system.
The remainder of this paper is structured as follows: Section 2 provides
some background on the existing command-oriented instrumentation
methodology. Section 3 covers the design considerations for model-driven
instrumentation derived from command-oriented instrumentation systems
while Section 4 describes key transition implementation experiences. Section
5 covers work in related areas of interest and Section 6 concludes the paper.
2 Background
Prior to discussing the transition challenges and experiences, for a better
understanding of the problem domain, a brief overview of command-oriented
instrumentation methodology is provided.
Management Instrumentation interfaces constructed around command-
orientation do not typically adhere to crisp layered interface and object-
oriented data-hiding principles. Management instrumentation systems often
begin with the simplest management method and grow as requirements grow,
starting with support of feature-based commands targeted for particular
command-line driven instrumentation needs. These CLI interfaces involve
the parsing of a user-entered command that link to a monolithic action
function. These action functions perform validation of system and feature
state and either configure some data or display a hard-coded report describing