Электронная библиотека БГУ: Главная страница

A new paradigm for the design of high-performance computing systems is emerging. The new paradigm is based on the concept of data-intensive computing, where the data processing is the main bottleneck. In traditional computing systems, the processing of instructions is the main bottleneck. In data-intensive computing systems, large amounts of data are generated or collected, and they need to be processed efficiently to extract useful information. This requires new architectures, new hardware, and new software approaches.

One approach to data-intensive computing is called in-memory computing. In in-memory computing, the main memory is used as the primary storage for both data and instructions, instead of the traditional separation between main memory and secondary storage. This allows for much faster data access times and reduces the need for data transfer between different levels of memory and storage.

Another approach to data-intensive computing is called parallel computing. Parallel computing involves dividing a large problem into smaller sub-problems that can be solved simultaneously on multiple processors or cores. This can significantly reduce the time required to process large amounts of data.

A third approach to data-intensive computing is called accelerator-based computing. Accelerator-based computing involves using specialized hardware accelerators, such as GPUs or FPGAs, to perform specific data processing tasks much faster than traditional CPUs. This can be particularly useful for data-intensive applications that involve large amounts of data processing, such as scientific simulations or machine learning.

These approaches can be combined to create high-performance computing systems that are optimized for data-intensive workloads. For example, an in-memory computing system can be combined with parallel computing and accelerator-based computing to create a system that can efficiently process large amounts of data