Integrated Modeling Tool for Indexing and Analyzing State Machine Trace

Catalog abstract summary

The accessible abstract describes a modeled state representation with a special-purpose database, filtering/highlighting, and Gantt/XY views for Linux and Windows traces.

Source: OpenAlex abstract metadata and institutional metadata reviewed; abstract not reproduced because reuse permission was not established.

Problem and motivation

Comprehensive traces are large and domain-specific, while conventional analysis tools can be hard-coded and custom XML/DSL analyses are difficult to author and maintain. Analysts need a declarative way to describe states, conditions, and event-driven transitions for kernel and user-space traces.

Method and contribution

The tool implements a UML state-machine subset with Graphiti and EMF and integrates as a Trace Compass plug-in. Analysts define attributes in a tree, states and transitions, Boolean AND/OR/NOT conditions, and event-to-state mappings. The model is transformed into a Trace Compass XML state provider in the current implementation; state history is stored in the disk-based State History Tree and generated views expose the result. The evaluated tracing pipeline uses LTTng and LTTng-UST on Linux; SystemTap, perf, and DTrace are discussed as related tracing ecosystems rather than the evaluated collector.

Findings and evidence

Two use cases are demonstrated: a Linux-kernel model with 14 states, 36 transitions, and 9 conditions, and a user-space request model with 7 states, 11 transitions, and 5 conditions. The corresponding attribute counts are 594 versus 525 for the kernel model/XML and 261 versus 196 for the request model/XML. The graphical workflow reproduces the XML analysis result without recompiling a Java analysis for each requirement and supports the same analysis style for different trace types.

Limitations and future directions

Limitations: The current implementation still compiles/transforms the graphical model to XML, so it does not yet remove every intermediate representation. The case studies demonstrate expressiveness rather than broad usability or performance; there is no large user study or cross-platform benchmark.

Future work: Execute models directly, add model validation and event-sequence checking, support model/repository management and trace-to-model links, and benchmark authoring effort, query cost, and analyst correctness across larger kernel and application analyses.

Sources and identifiers

• Published version published
• PolyPublie institutional publication record public_source_record

When to cite this paper

Cite this paper when specifying Trace Compass analyses through graphical state-machine models over LTTng traces.

• Graphiti/EMF state-machine modeling integrated with Trace Compass.
• Declarative states, transitions, Boolean conditions, and event mappings.
• The kernel and user-space request case studies and model sizes.
• The current XML transformation boundary and planned direct execution.

Citation

@inproceedings{ezzatiJivan2021integratedmodeling,
  author = {Simon Delisle and Naser Ezzati-Jivan and Michel R. Dagenais},
  title = {Integrated Modeling Tool for Indexing and Analyzing State Machine Trace},
  year = {2021},
  booktitle = {2021 IEEE International Symposium on Networks, Computers and Communications (ISNCC)},
  pages = {1-8},
  publisher = {IEEE},
  doi = {10.1109/ISNCC52172.2021.9615814},
  url = {https://doi.org/10.1109/ISNCC52172.2021.9615814}
}

Delisle, S., Ezzati-Jivan, N., & Dagenais, M. R. (2021). Integrated Modeling Tool for Indexing and Analyzing State Machine Trace. In 2021 IEEE International Symposium on Networks, Computers and Communications (ISNCC) (pp. 1-8). https://doi.org/10.1109/ISNCC52172.2021.9615814

S. Delisle, N. Ezzati-Jivan, and M. R. Dagenais, "Integrated Modeling Tool for Indexing and Analyzing State Machine Trace," in 2021 IEEE International Symposium on Networks, Computers and Communications (ISNCC), pp. 1-8, 2021, doi: 10.1109/ISNCC52172.2021.9615814