The executable model approach: synergistic strategies for maintaining the correctness and efficiency of computer networks
OA Version
Citation
Abstract
The Executable Model Approach offers an effective way to integrate performance
modeling into the entire development cycle of computer networks. An executable performance
model is created early in the development cycle and continues to be refined in
parallel with network software. The model has special system requirements that make it
a fast, inexpensive, and accurate predictor of network behavior. This predictive capability
enhances Capacity Management tasks for both the design and operation of networks.
The Executable Model Approach is a set of strategies that use an executable performance
model to increase the correctness and efficiency of networks. The model can be
used as a testbed to confirm hypotheses about network behavior and to speed up the process
of troubleshooting performance faults. The model also represents an independently
developed encoding of network algorithms and can be used in the fashion of N-Version
programming to increase the reliability of network software. Finally, the model is instrumented
with finer detail than the operational network and can be used to enhance the
understanding of network behavior.
Executable models have different properties than models traditionally used to design
networks. Executable models are optimized for fast turn-around time and their functionality
is limited to modeling a specific operational context, whereas design models have
flexible functionality that allows them to explore a wide range of hypothetical networks.
The mechanisms needed to create and run executable performance models were not available until the recent advances in expert systems, supercomputers, and hybrid-modeling.
The feasibility of creating executable models is demonstrated by the development of
an automated tool for troubleshooting performance faults. Also, the potential benefit
of using executable models is analyzed in terms of increased software reliability and
troubleshooting speed.
The Executable Model Approach is effective because it exploits the characteristics of
both computer networks and executable models. Computer networks are large, complex,
and expensive systems. Methods are needed to analyze their behavior in a general
and global way. Executable models automatically generate a representation of network
behavior from network stimulus. They are easy to use and combine with other tools.
Description
License
This work is being made available in OpenBU by permission of its author, and is available for research purposes only. All rights are reserved to the author.