Abstract: An embodiment may involve persistent storage containing definitions of a set of queues and a set of servers, and wherein the servers are respectively associated with deactivation times. One or more processors may be configured to: (i) identify, by an assignment engine, an incoming request in a particular queue, wherein the incoming request is ready for assignment to one of the servers; (ii) determine, by the assignment engine and based on data associated with the incoming request, an expected duration for servicing of the incoming request; (iii) calculate, by the assignment engine and based on the deactivation times, times remaining in service for each of the servers; (iv) select, by the assignment engine and from the servers, a particular server that has a time remaining that is greater than the expected duration; and (v) assign, by the assignment engine, the incoming request to the particular server.