This article deals with a simple optimization for a level-5 protocol called JRMP (Java's Remote Method Protocol), which is used in a distribution model named Java's RMI (Java's Remote Method Invocation). The main JRMP subprotocol, namely Stream, has been enhanced with a simple and direct multiplexing mechanism that offers the possibility of transferring several parallel request-response interactions without opening new TCP/IP connections. The overhead required to process headers and the advantages stemmed from the approach in terms of response-time are explored on a switched-ethernet benchmark application. Este trabajo ha sido financiado parcialmente por el proyecto nacional REM4VSS (TIN 2011-28339) y eMadrid (S2013/ICE-2715) y HERMES- SMART-DRIVER(TIN2013- 46801-C4-2-R)