Using traceability in model-to-model transformation to quantify confidence based on previous history

Saxon, John Trevor (2018). Using traceability in model-to-model transformation to quantify confidence based on previous history. University of Birmingham. Ph.D.

Preview

Saxon18PhD.pdf
PDF - Accepted Version
Download (1MB)

Abstract

A widely used method when generating code for the purposes oftransitioning systems, security, the automotive industry and other mission critical scenarios is model-to-model transformation. Traceability is a mechanism for relating the source model elements and the destination elements. It is used to identity how the latter came from the former as well as indicating when and in what order. In these application domains, traceability is a very useful tool for debugging, testing and performance tuning of model transformations. Recent advances in big data technologies have made it possible to produce a history of these executions. In this thesis, we present a method on how we can use such historical data that quantifies the confidence a user has on a newly proposed transformation. For a given trace of execution, considering historical traces that are either well tested, or performed correctly over time, we introduce a measure of confidence for the new trace. This metric is made to compliment that of traditional testing and verification. For example, our metric will aid in deciding whether to deploy automatically generated code when there is not enough time or resources for thorough testing and verification. We shall evaluate our framework by providing a transformation that transitions a relational database into that of a NoSQL database, specifically Apache HBase. This transformation involves changing the nature of the data that is mapped, such that a loss in integrity occurs in the event of its failure.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):