Improving FFDA of Web Servers through a Rule-Based Logging Approach

Log production is known to be a developerdependent, error-prone process. This results in heterogeneous and inaccurate logs, which can mislead Field Failure Data Analysis (FFDA). Research papers face logs issues by proposing manipulation techniques to be applied to the field data source as it is. This source can however miss important log entries, decreasing the level of trust on FFDA results. This paper describes how to improve the field data production by means of a rule-based logging approach preventing inaccurate logs. It also presents an experimental evaluation with reference to the widely used Apache Web Server. Results show that the rule-based approach significantly improves exiting Web Server logs through a better coverage of timing failures. Keywords-Field Failure Data Analysis, Logging Rules, Fault Injection, Web Server.