Students delve into several of the newest topics of the constitution method through key articles, with a focus on supercomputers. The largest scale modern supercomputers have several million CPU cores, and internal parallelism from several million to 10's of millions through hardware multithreading. In addition, to provide them with data, there is a petabyte/seconds level memory system, and the whole is connected by a network with performance that rivals the internet as a whole. By discussing these technologies in class, the constituent factors are made clear, and students grasp the technological essence necessary for large-scale computing.
The objective is for the students to gain deep insights into some of the latest key technologies that enable modern supercomputers to function, such as resilience, massive (big) data. Many core / GPU technologies. Such technologies as well as algorithms and programming to enable their usage is not only restricted to supercomputing, but will be the key enablers for large-scale computing in IT systems in general.
Supercomputer, Supercomputing, HPC, many-core / GPU, big data, fast interconnect. High bandwidth memory, parallel processing, parallel programming languages, Resilience, Low power computing
✔ Specialist skills | ✔ Intercultural skills | ✔ Communication skills | ✔ Critical thinking skills | Practical and/or problem-solving skills |
The class will be taught in flip teaching style + active discussions in “US University” classroom style. A designated student will select the topical paper and present in front of the entire class and required to answer to the questions. Other students will also pre-read the papers and will join in the active Q&A and discussions. The result is deep understanding of the subject matter by all participating members of the lecture.
Course schedule | Required learning | |
---|---|---|
Class 1 | History of Supercomputing | none |
Class 2 | Leading applications in supercomputing | required reading of the paper before class |
Class 3 | hardware architecture in supercomputers: CPUs | required reading of the paper before class |
Class 4 | hardware architecture in supercomputers: interconnect | required reading of the paper before class |
Class 5 | hardware architecture in supercomputers: high bandwidth memory | required reading of the paper before class |
Class 6 | big data handling in supercomputers | required reading of the paper before class |
Class 7 | system software in supercomputers: OS | required reading of the paper before class |
Class 8 | system software in supercomputers: runtime | required reading of the paper before class |
Class 9 | parallel programming in supercomputers: OpenMMP | required reading of the paper before class |
Class 10 | parallel programming in supercomputers: MPI | required reading of the paper before class |
Class 11 | performance modeling in supercomputers: basics | required reading of the paper before class |
Class 12 | performance modeling in supercomputers: performance tools | required reading of the paper before class |
Class 13 | resilience in supercomputers: detection and checkpointing | required reading of the paper before class |
Class 14 | resilience in supercomputers: recovery | required reading of the paper before class |
Class 15 | Future of supercomputers | required reading of the paper before class |
none
Oyanagi, Sato, Nakamura, Matsuoka. "Supercomputer". Iwanami-Shoten (Iwanami Publishing), 2012 (in Japanese)
Presentation during class (40%), Q&A during class (30%), and the final report (30%).
Basic knowledge of computer architectures and parallel processing, C Language