Visual Analytics for Large Communication Trace Data

Executions of modern parallel programs often yield complex communications among compute nodes of large-scale clusters of workstations or supercomputers. Analyzing communication patterns is becoming increasingly critical to performance optimization. As the scale and complexity of parallel applications drastically increases, visu-alization has become a feasible means to conduct analysis of massive communication patterns. However, most visualization tools fall short in showing comprehensive dynamic communication graph and addressing the scalability issue. Our solution for analyzing dynamic communication patterns is based on an analytics framework coupled with a new visualization technique, named CommGram [29], that provides a flexible solution to the scalability issue. We can explore large communication data at different levels of detail, and detect potential communication bottlenecks of massive parallel programs. The conclusion of our studies is based on large-scale scientific applications that include end-to-end simulation pipelines and AMR-based simulations. iii ACKNOWLEDGMENTS Professor Hongfeng, my advisor, is on top of my appreciation and acknowledgement list. Through his endless support, guidance, feedback and critique throughout my graduate work, I have achieved many of my goals and also developed a broad set of skills ranging from scientific thinking, to programming and data analysis.