Challenges and Experiences in Transitioning Management Instrumentation from Command-Oriented to Model-Driven - Advances in Network-Embedded Management and Applications - page 8

through a common abstract interface design. This abstract middleware
interface directs instrumentation access requests to appropriate
instrumentation for manipulation of data and services through their
instrumentation API.
The instrumentation API is collection of functions that independently
access specific instrumentation data and services. The framework of
functions may be generated from information in the instrumentation model,
but not the particular code to access the actual data or service of the
instrumentation - that must be supplied by the instrumentation developer.
When implementing APIs for modeled legacy instrumentation, there are
two potential sources for instrumentation API implementation source code: 1)
port it from the legacy implementation, 2) write it from scratch. It is
important to consider which method is the most accurate, reliable, and most
reusable. Frequently, only fragments of command-oriented instrumentation
source code may be leveraged in a model-driven design. A careful evaluation
of the effort required to port existing code or to write new and perhaps more
efficient instrumentation code should be carefully considered.
Creating model-driven instrumentation can be more challenging than
command-oriented instrumentation development if model-imposed
restrictions are to be used. Command-oriented development allows direct and
freeform access to feature instrumentation at anytime, from virtually
anywhere, with no interface definition requirements. In contrast, because of
crisply defined constructs, modeled instrumentation has the capability to
ensure rigorous data validation and consistent instrumentation interfaces
between clients and available instrumentation data and services. While this
makes API definition somewhat more challenging than the freeform method,
this is one of the premier benefits of model-driven systems and results in
improved reliability and quality.
4 Implementation Experiences
This section will explore the challenges and experiences gained through the
prism of a real-world development effort where an established command-
oriented management instrumentation system was transitioned to a model-
driven management instrumentation system. It will discuss design choices
and the reasoning behind them.
The most design-influential aspect of the transition was resolving the