The article discusses the tasks of forming and scheduling the deployment of IT service releases, the settings of which are based on the data about the structure of services stored in the configuration database (CMDB) of the service provider and that help to estimate the number of IT service failures caused by violations in the state of the operational environment due to release deployment. Three sequentially solved problems are considered: the problem of determining the number of releases, the task of determining the releases composition and the task of determining the schedule for deploying releases. The proposed mechanism for determining the number of releases minimizes the deployment costs and limits the failure rate of IT services by the capacity of a channel servicing the failures. The problem of forming the releases composition is formulated and solved as a problem of stones. The proposed mechanism for determining the release deployment schedule uses the discrete version of the local optimization method and is aimed at minimizing the average failure service time while limiting the "lost time" (time duration of incomplete load of the service channel, which increases the total processing time of all failures).