A method, an apparatus, and a system for transmitting media data based on OTT are disclosed. The method includes: receiving, by an OTT server, a media data request initiated by a user terminal, where the OTT server is connected to at least one cache server; determining, by the OTT server, a cache server from the at least one cache server according to a location and/or a capability of the at least one cache server; sending, by the OTT server, a notification message, which carries session information, to the user terminal and the determined cache server, so that the determined cache server receives a media data request, which carries session information, from the user terminal, and sends corresponding media data to the user terminal.