Cronista: A multi-database automated provenance collection system for runtime-models

Owen Reynolds*, Antonio García-Domínguez, Nelly Bencomo

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

Context:
Decision making by software systems that face uncertainty needs tracing to support understandability, as accountability is crucial. While logging has been essential to support explanations and understandability of behaviour, several issues still persist, such as the high cost for managing large logs, not knowing what to log, and the inability of logging techniques to relate events to each other or to specific occurrences of high-level activities in the system.

Objective:
Cronista is an alternative to logging for systems that act on top of runtime models. Instead of targeting the running systems, Cronista automatically collects the provenance of changes made to the runtime models, which aim at leveraging high-level representations, to produce more concise historical records. The provenance graphs capture causal links between those changes and the activities of the system, which are used to investigate issues.

Method:
Cronista’s architecture is described with the current design and the implementation of its high-level components for single-machine, multi-threaded systems. Cronista is applied to a traffic-simulation case study. The trade-offs of two different storage solutions are evaluated, i.e. the CDO model repositories, and JanusGraph graph databases.

Results:
Integrating Cronista into the case study requires only minor code changes. Cronista collected provenance graphs for the simulations as they ran, using both CDO and JanusGraph. Both solutions highlighted the cause of a seeded defect in the system. For the longer executions, both CDO and JanusGraph showed negligible overhead on the simulation times. Querying and visualisation tools were more user-friendly in JanusGraph than in CDO.

Conclusion:
Cronista demonstrates the feasibility of recording fine-grained provenance for the evolution of runtime models, while using it to investigate issues. User convenience and resource requirements need to be balanced. The paper present how the available technologies studied offer different trade-offs to satisfy the balance required.
Original languageEnglish
Article number106694
JournalInformation and Software Technology
Volume141
Early online date8 Aug 2021
DOIs
Publication statusPublished - Jan 2022

Bibliographical note

© 2021, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0

Funding: The work was partially funded by the Leverhulme Trust Research Fellowship (Grant RF-2019-548) and the EPSRC Research Project Twenty20Insight (Grant EP/T017627/1).

Keywords

  • Multi-threading
  • Provenance
  • Runtime-models
  • Self-adaptation
  • Self-explanation

Fingerprint

Dive into the research topics of 'Cronista: A multi-database automated provenance collection system for runtime-models'. Together they form a unique fingerprint.

Cite this