Multimedia real-time session services, such as voice and videoconferences are highly desirable in many mobile ad hoc networks (MANET). This paper investigates network architecture options and evaluates their performance metrics for supporting session-based multimedia services in MANET, using the standard Session Initiation Protocol (SIP) proposed by the Internet Engineering Task Force (IETF). A novel service discovery mechanism is presented to discover SIP service locations and capabilities, enabling service self-configuration in the MANET. The service discovery mechanism applies cross-layer design between the application and the routing layers to achieve improved system and protocol efficiency. Two SIP MANET networking architectures, both with and without the proxy server, are evaluated. The presented performance results demonstrate the efficacy of the solution, and illustrate network design guidance for supporting SIP applications over MANET