Building Resource Adaptive Software Systems - Phase 2 Evaluation BAE: Regenerative INtent-Guided Systems (RINGS)
Abstract
In phase 1 RINGS was developed, which provided a general way to augment existing applications with the goal of making them more resilient to changes within their operating environment. This was achieved through the development of a monitoring architecture which detected functional deviations from intent as specified in the FIDL specification language. When the monitors detect conditions requiring repair, a genetic search algorithm was triggered, guided by a fitness function. If a suitable modification was generated, the application was redeployed with that change incorporated. The three challenge problems in phase 2 extend on this previous work to explore both functional and non-functional application intent. In addition, a new approach to developing intent preserving applications via correct-by-construction formal specification is presented. Challenge problem 1 ATIF Auto-Focus demonstrates RINGS genetic algorithms adapting the configuration parameters to produce an improved detection picture of the observed tracks. Challenge problem 2 Legacy Optimization and Repair is a continuation of the phase 1 work to repair message parsing code in the presence of perturbations to the incoming message format and adds a degree of non-functional intent by placing constraints on the amount of system memory the tracker is allowed to utilize. Finally, challenge problem 3 Adapting Formal Specifications focuses on adapting formal specifications to address both functional adaptation and refinement of the implementation to optimize for resource utilization.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 05, 2018
- Accession Number
- AD1109553
Entities
People
- Eric Kilmer
- Jeffrey Hughes
- Martine Kalke
- Philip Zucker
- Robert Cunningham
- Timothy Braje
- Timothy Meunier
- W. Konrad Vesey
Organizations
- MIT Lincoln Laboratory