Meta-analysis of publications on software logging

Abstract

Logs are widely used to monitor software systems. They provide rich information to assess system runtime behavior and detect anomalies. Log generation and significance have been studied in many publications. The multiplicity of appearing publications hampers tracing research advances. Review papers on software logging address these problems. Nevertheless, the problem of publication selection tailored to the readers’ interest still arises. To tackle this, we have proposed combining a publication taxonomy with a meta-analysis of derived characteristic features. This approach has been enhanced by the development of specialized analysis tools based on natural language processing. These tools can also be used for analyzing publications in other domains.