Platform development for parallel operation of single board computers

Recent improvements in technology like mobile devices, Internet of things equipment or scientific and industrial applications generated large amounts of data and necessity of high performance hardware to process this data. Besides, because technological improvements, the processing of large programs, the transformation of enormous data, and the increase of systems which have the simultaneous data flow from interconnected discrete elements have become an inevitable part of daily life. Even though there exist expensive servers there is always a need for all these needed are required fast, affordable, scalable, efficient and flexible solutions. At this point, High-performance computing and parallel programming can meet at this bottleneck to solve data-rich, large-scale programs. Beowulf clustering is one of the high-performance computing approaches which is provides multiprocessing platform to run programs into as divided tasks parts concurrently. In this thesis, a scalable Beowulf cluster consisting of single board computers (SBC) were built and were evaluated its performance is evaluated. Single board computers are preferred as a computing node because of its credit card size, affordability and its ability to meet to the required performance that is needed. The infrastructure operating system of the platform was developed based on Linux operating system. Python was selected preferred as the development environment programming language and MPI was used to the (message passing interface) and MPI4py were used to carry out parallel operations. The platform was tested with problems programs with different characteristics and results were evaluated compared according to with previous studies in this field. Scalable speed up tendency is observed from test results.