PreX: A Predictive Model to Prevent Exceptions

Authors

Abstract

The exception handling mechanism has been one of the most used reliability tools in programming languages in the last decades. However, this model has not changed much with time, in spite of advances in programming languages, which include concurrent programming and a shift towards more reactive paradigms, the basic principle remains the same - an exception occurs, and the mechanism reacts. We propose a new paradigm, inspired by Online Failure Prediction (OFP), to predict exceptions and possibly avert them by triggering the execution of preventive actions. The proposed model - PreX - is, thus, proactive, operating in a much finer-grained level than any other form of online failure prediction. OFP has shown promising results in predicting failures at a higher level, but has never been available to the developer, being mainly a system level technique. Thus, PreX will offer developers a new range of revitalization strategies. In this work, we describe the model and evaluate its performance by applying it to a real e-commerce solution, demonstrating how it is capable of predicting and preventing exceptions at run-time. Furthermore, we also show that PreX increases the overall availability and performance of the system under the same conditions.

Keywords

Exception handling, Proactive, Failure prediction, Self-healing, Preventive, Predictive

Subject

PreX: A Predictive Model to Prevent Exceptions

Journal

Journal of Systems and Software, Vol. 137, pp. 652-668, Elsevier, March 2018

DOI