Engineers and scientists in the supercomputing field will immensely benefit from this book. For software engineers or any computing practitioner, the book presents up-to-date projects, formats, and libraries for all levels of the input/output (I/O) stack. The levels range from the bottom of the stack, with parallel file systems and hardware architecture details, to the top of the stack, with high-level I/O libraries and applications. For data scientists, the book provides a look into major supercomputing centers in the US, highlighting hardware and software decisions they made, and the main challenges in data organization and data analysis they face. The book also has the potential to be central in many industries, not only in the niche of computational sciences, as big data problems arise and the Internet of Things becomes a reality.

The book is a compendium of topics within the scope of high-performance parallel I/O, with each chapter written by different authors. The editors performed the extraordinary feat of selecting appropriate topics and documenting the technological progress accomplished in the fast-evolving area of high-performance computing. It is also remarkable to have a broad overview of past, present, and future technologies, with a focus on real problems and clear solutions rather than misleading, imaginary cases. Two examples of recurring discussions in the book include bandwidth throughput and performance measurements over a usage period, instead of appealing to disk capacity and ideal peak performance. The book is divided into six self-contained parts, wherein several chapters enumerate the topics of interest.

The first part of the book immerses the reader in the practice of parallel I/O carried out in existing supercomputing centers, and it wonderfully succeeds, given the hands-on experience of the authors and editors. The second part of the book describes four parallel file systems, two open-source projects, and two software products, giving the reader an important perspective on the backbone of a high-performance computing facility. The third part of the book details six promising I/O libraries, especially the excellent Hierarchical Data Format version 5 (HDF5) library, encompassing different levels of the I/O stack, with interesting historical background and practical usage directions. The fourth part lists case studies on parallel I/O, and nicely complements previous parts of the book, as each chapter provides the hardware and I/O library choices and discussions on why these choices are made. The fifth part of the book specifies essential I/O profiling tools, helping the reader navigate the differences in components, workloads, and scheduling policies of the underlying I/O subsystem.

The sixth and final part auspiciously closes the book, giving future directions and insights in the turbulent area of high-performance computing. Technical directions include burst buffers and in-transit analysis, and an example of a promising future technology is silicon photonics, due to the increased bandwidth and reduced power and complexity. These future trends could be the foundation for scaling up to exascale platforms.