A method, implemented in a distributed system including a plurality of processing devices, generates an end-to-end Call Detail Record (CDR) on voice and multimedia telecommunication sessions over a plurality of telecommunication network elements and interfaces. The method includes the steps of receiving CDRs generated in real-time and streamed by multiple network elements, each CDR referring to a specific interface or protocol, processing the received CDRs as they are received to identify the specific interface or protocol to which the received CDR refers and to identify a single telecommunication session based on a key associated with the received CDR, and creating an end-to-end CDR incrementally based on all of the received CDRs identified for the single telecommunication session, wherein at least two of the received CDRs have a different interface or protocol from one another.