Design Guidelines and Visualization Support for Developing Parallel Real-Time Applications

Large telecommunication applications running in real-time environments need to have high, predictable and scalable performance. Parallel (multithreaded) programs and multiprocessors is one way to achieve this. However, the experience of writing parallel programs is still limited and a lot of issues need to be addressed. Moreover, tools and methods are needed to predict and visualize the performance of a parallel program when using a multiprocessor, e.g. how will performance scale when the number of processors is increased, and where are the serialization bottlenecks. As a part of this project we will extend an early prototype of a performance prediction and visualization tool. By studying a set of real telecommunication applications, provided by our industrial partner Ericsson Software Technology, we will also obtain a set of guidelines for writing parallel programs with high, predictable and scalable performance. These guidelines will also consider maintainability and reusability aspects. The collaboration with the academic partners within PAMP will make it possible to use the same tools for performing the research. Our research will be at the application level, and will thus complement the research at the operating system and hardware level performed by other nodes in PAMP. Although Ericsson Software Technology is our main industrial partner, we hope that we will be able to use applications from the other industrial partners as well.