Raspberry Pi is a single-board computer developed by a British charity to promote a different way of teaching computer science in schools, away from applications and web design and geared toward programming and experimentation. Raspberry Pi is about the size of a credit card, peripherals like monitor and keyboard are not included, and it is rather inexpensive. It became far more popular than anticipated and is now used not only in schools but in other environments as well.
The term supercomputer dates back to the early 1960s and was in vogue throughout the 1980s. It denotes specialized high-performance computers used for a range of computationally intensive tasks in various fields, like weather forecasting, oil exploration, astrophysics, cryptanalysis, and so on. From the 1990s on, the terms computer cluster or massively parallel computers were used to reflect a paradigm change toward an architecture composed of a large number of interconnected general-purpose computers.
One would think that Raspberry Pi and supercomputing are not uttered in the same breath, but in fact the small low-performance computer has already been used to build clusters of several thousand computing cores. For example, in 2017 a Raspberry Pi cluster of 3,000 cores was built at the Los Alamos National Laboratory to simulate large-scale computer clusters for exascale computing, that is, systems with a performance 50 times higher than today’s supercomputers [1]. Well, the supercomputing ambitions described in this book are far more modest. About 40 pages (about one quarter of the book) are devoted to this topic. The author describes the message-passing interface MPI4PY used with the Python programming language to pass information between different processors. With the MPI4PY software installed, several Raspberry Pis can be interconnected via a Wi-Fi router. This is described in some detail in a separate chapter, listing all necessary configuration commands. Having a small cluster of four Raspberry Pis set up, the performance can be further enhanced by overclocking the processors, that is, by increasing the processors’ clock frequency. A hardware cluster as such does not yield a higher performance, if the software does not take advantage of the increased processing possibilities. In ten pages, parallel programming in Python is very briefly described.
The first part of the book provides an introduction to the setup of the Raspberry Pi, explains useful Linux commands, and shows how to connect from a PC to the Raspberry Pi. It also lists some features of the Python programming language. The last five chapters provide an overview of Python software libraries that can be used for scientific programming, namely SciPy, NumPy, and Matplotlib. Two applications for those libraries are described in more detail--signal processing and image processing. The part on scientific programming is not directly related to the part on supercomputing. That is very unfortunate, because one major field of application for supercomputers is scientific programs. So the book leaves the reader with a rather scattered impression. It covers too many topics without relating them properly and without always providing sufficient details. However, as the subtitle says, the book is for enthusiasts, who will enjoy it and dig further into the topics by accessing further resources.
More reviews about this item: Amazon
