Cloud applications monitoring: An industrial study

Abstract

Modern software systems employ large IT infrastructures hosted in on-premise clouds or using “rented” cloud resources from specific vendors. The unifying force across any cloud strategy is incremental product and application improvement against conservation of those resources. This is where monitoring of cloud applications becomes a key asset To shed light over the status of monitoring practices in industry, we study: (a) monitoring practices and tools adoption in industry; (b) size and complexity of industrial monitoring problems; (c) the role of software architecture and software process with respect to monitoring strategies. We conduct mixed-methods empirical research featuring interviews and a web survey featuring 140+ practitioners from over 70 different organizations. Even if the market makes available a significant set of monitoring tools, our results show a rather unappealing picture of industrial monitoring: (a) industrial decision-makers do not perceive monitoring as a key asset even though the downtime of their applications correlates heavily with the level of automation and responsiveness enabled by monitoring; (b) monitoring is done with crude technology, mostly MySQL querying or similar (e.g., Nagios); finally, (c) incidents are discovered by clients rather than application owners. We conclude that the road toward the industrial adoption of cutting-edge monitoring technology is still one of the less travelled, presumably in connection to the considerable investment required. Furthermore, the lack of industrial cloud monitoring standards does not help in addressing the proliferation of multiple tool combinations, with varying effectiveness. Further research should be invested in looking into and addressing these major concerns.