Building Affordable High Performance Computing Platforms for Engineering Education

Traditionally, engineering and science disciplines have relied on observation, theory, and experimentation as tools to perform research to explore new knowledge.With the introduction of computer hardware and software, numerical simulation based on mathematical modeling gradually becomes an important tool.After high performance computers are mature and commercially available, numerical simulation has become a tool as important as observation, theory, and experimentation to all engineering and science disciplines.In most cases, it is adopted more often than experimentation because it is more economic, less time-consuming, and able to explore infeasible situations.Without doubts, numerical simulation should be included by engineering education.Numerical simulation depends on high performance computers.High performance computers refer to parallel computers, namely computers equipped with multiple processors.However, the largest barrier to high performance computing education is the high cost of high performance computers.Most institutions cannot afford expensive parallel computers.The author explored and managed to discover three options of affordable platforms suitable for high performance computing classes.The first platform is personal computers equipped with multi-core processors and thread libraries.The second platform is PCs equipped with Graphic Processing Unit cards.The third platform is commodity clusters consisting of inexpensive PCs and network switches.They even can be built with retired PCs.All three approaches provide low-cost solutions for all institutions to offer their high-performance computing education.