Implement of a high-performance computing system for parallel processing of scientific applications and the teaching of multicore and parallel programming

Abstract

[EN] Increasingly complex algorithms for the modeling and resolution of different problems, which are currently facing humanity, has made it necessary the advent of new data processing requirements and the consequent implementation of high performance computing systems; but due to the high economic cost of this type of equipment and considering that an education institution cannot acquire, it is necessary to develop and implement computable architectures that are economical and scalable in their construction, such as heterogeneous distributed computing systems, constituted by several clustering of multicore processing elements with shared and distributed memory systems. This paper presents the analysis, design and implementation of a high-performance computing system called Liebres InTELigentes, whose purpose is the design and execution of intrinsically parallel algorithms, which require high amounts of storage and excessive processing times. The proposed computer system is constituted by conventional computing equipment (desktop computers, lap top equipment and servers), linked by a high-speed network. The main objective of this research is to build technology for the purposes of scientific and educational research.