Evaluation of the Performance of a Computing Cluster Based on Single Board Raspberry Pi 3B + Computer
Olesia Barkovska, Valentyn Korniienko, Инна Викторовна Филиппенко, Vladyslav Adamovych, Oleh Filippenko · 2023
The article is dedicated to the problem of parallel computing on multicomputer accelerated data processing systems, namely on cluster computing systems, which consist of many independent computers and the network that connects them. The computing components of the cluster being designed are four single-board Raspberry Pi 3B+ computers. Raspbian Buster is used as the operating system for the designed cluster. The cluster system uses a SD memory card as a disk array and does not have the ability to connect nodes to the only one drive, so the topology of cluster pairs uses a distributed access topology that allows each memory disk to connect to only one cluster node. Messaging between processes is provided by the MPI interface. One of the main advantages of the cluster is the ability to scale, which, in the case of increasing computational load, leads to increased acceleration and efficiency of parallel computing. The performance of the created computing cluster based on four single-board Raspberry Pi 3B+ computers was tested on the problem implementation of the large matrices accelerated multiplication, as matrix operations are the basis of scientific and engineering calculations, and are used in mathematical modeling of various processes, phenomena and systems.