Structured and Interoperable Logging for the Cloud Computing Era: The Pitfalls and Benefits

Computer logs represent a valuable source of monitoring information being typically the only mechanism for gaining visibility into the behavior of monitored resources. However, even in modern cloud data centers the majority of logs is produced in an unstructured data format with free-form natural language messages carrying the most important portion of the information. This is source of many issues that prevent logs from being effectively and intelligently processed in an automated way. We argue that it is an essential flaw when considering the volume and heterogeneity of logging data produced by cloud. This paper elaborates on the common problems and shortcomings in the domain of (application) logging, and presents an experimental logging mechanism for Java language that is capable of producing logs in a unified and extensible format allowing for their efficient automated processing.