This paper discusses a model-based architectural approach for improving predictability of performance in embedded real-time systems. This approach utilizes automated analysis of task and communication architectures to provide insight into schedulability and reliability during design. Automatic generation of a runtime executive that performs task dispatching and inter-task communication eliminates manual coding errors and results in a system that satisfies the specified execution behavior. The MetaH language and toolset supports this model-based approach. MetaH has been used by the U.S. Army in a pilot project applied to missile guidance systems. Reduced time and cost benefits that have been observed will be discussed as a case study. The paper closes by outlining the current state of commercial availability of such technology and efforts to develop standards, such as those put forth by the Society of Automotive Engineers (SAE); Avionics Systems Division (ASD); working group on Avionics Architecture Description Language (AADL); and the Object Management Group (OMG) Unified Modeling Language (UML) working group on real-time and performance support in UML.