Large Scale Hybrid Computing
Objectives
- Characterise the architecture of large-scale supercomputers, including the key components and their interconnections
- Evaluate the impact of architectural choices on the scalability of parallel applications on heterogeneous servers
- Develop scalable parallelisation strategies for applications running across multiple servers
- Assess appropriate data structures and load-balancing strategies for efficient large-scale computation
- Identify common patterns and best practices in the development and optimisation of exascale applications
- Evaluate the potential of emergent technologies on large-scale architectures
Program
- Architecture of large-scale supercomputers
- Inter-server interconnection fabrics and topologies at scale
- Hybrid parallelisation on heterogeneous servers
- Scalable multi-server parallelisation
- Multi-level workload scheduling
- Data structures and domain decomposition techniques
- Performance assessment of large-scale applications
- Overview of successful exascale applications
Bibliography
- Computer Architecture: a Quantitative Approach, David Patterson and John Hennessy, Morgan Kaufmann, 2017
- Supercomputer Architecture, Paul Schneck, Springer, 1988
- Large-Scale Computing Techniques for Complex System Simulations, Werner Dubitzky and Krzysztof Kurowski and Bernard Schott, Wiley, 2011
- High Performance Computing, Thomas Sterling, Maciej Brodowicz, Matthew Anderson, Morgan Kaufmann, 2017