OVERVIEW AND EXPERIENCE OF ADOPTING AFFORDABLE PARALLEL COMPUTING PLATFORMS FOR HIGH PERFORMANCE COMPUTING EDUCATION

Due to the fact that the development of computer uniprocessor has met the physical limitation and its clock speed can no longer be significantly pushed, the design of processor has shifted into the direction of multi-core processors. The adoption of parallel programming is mandatory to utilize the multiple cores of multi-core processors. The shift to multi-core processors also has significantly reduces the costs of parallel computers built on multiple multi-core processors. That further promotes the popularity of high performance computing (HPC) which relies on parallel computers. In last decade, computational applications have been widely expanded and adopted as high performance computers are more affordable than ever. Now computational applications are utilized in application and research in various fields such as Physics, Chemistry, Biology, Engineering, Analytics, and Finance. Therefore, parallel processing and programming courses should be taught by every computer-related academic department. However, parallel computers are still not affordable to all institutions because even the entry level parallel computers still cost tens of thousands U.S. Dollars. Luckily, owing to the development of hardware and open-source software, the author managed to discover ways to build parallel processing platforms with minimal costs. The author has used the two platforms in his HPC courses and seen their effectiveness. In this paper, the author shares his experience in incorporating building and utilizing the affordable parallel computing platforms in HPC education.