To the problem of identifying failures in the information technology infrastructure by monitoring and analyzing indirect data

Research aspects of the problem of identifying failures in complex information technology (IT) systems are discussed. A failure is understood as an abnormal mode of operation of the IT infrastructure, in which the specified functionality of the business processes supported by it is not provided but the existing means of monitoring the functioning of the IT infrastructure do not raise alarms. In such situations, the conclusion about the failure can be formed only by indirect data, in particular, by the reaction of users contacting the support service, etc. The tasks of building identification systems for such abnormal situations, the so-called downdetectors, are considered in the context of some research problems of modern artificial intelligence: intellectual analysis of natural language texts, identification of cause-and-effect relationships in the analyzed data, training on precedents in open subject areas, etc. The paper proposes directions of scientific research and formulation of tasks, the solution of which is necessary to significantly increase the efficiency of detecting failures using downdetector methods.