Getting StartedΒΆ
Note
This page and many of the others that are part of this documentation use terminology such as profiles, convolution, network analysis, etc. These terms are used without explanation in many places to simplify much of the documentation. Readers are encouraged to read through all of Background, Theory, and Results for a complete understanding of these terms as well as the rest of the theory and background which acts as the foundation for all of this work.
This repository code and its documentation covers a novel technique
for analyzing and predicting network performance of distributed
applications, called Precise Network Performance Prediction,
(
), which has been implemented into a design-time
analysis tool that analyzes application and system
models. Furthermore, a run-time library has been developed which
enables application code to generate network traffic that adheres to a
specified network profile. Metrics about this traffic are
automatically recorded for offline analysis. Furthermore the library
enables management of such networked applications by forcing them to
adhere to their profiles; in this way it acts as a lightweight layer
on top of the communications middleware layer used by the applications
on the system. Finally, anomalous traffic can be detected by
receivers and disabled through out-of-band (OOB) communication to the
sender-side middleware.
- Background, Theory, and Results : for readers interested in learning the theory behind
this analysis and implementation; covers all of the relevant
background, theory, and results pertaining to .
- Using the Code : for readers interested in using ; provides an
explanation the interfaces provided by the tools and walks through
some examples to demonstrate how they can be used.
- The Inner Workings : for readers interested in extending this work or learning more about the complete implementation of the theory; provides a launching point which directs them to the APIs that have been developed for interfacing between the code’s modules.