In May, it will be two years after the acquisition of the Bull Company by Atos, a financial transaction with the French company that expressed its desire to strengthen its presence in the fields of cloud computing, big data and security. As an expert brand in Atos technologies, software and hardware products, and the chosen device is part of the Bullx B520 series and is characterized by being a system capable of offering bespoke solutions based on the current industry standards. that allows to innovate in a much faster way. In addition, during its development, optimization and simplification have been taken into account for high performance computing or the commitment to a flexible structure for incoming and outgoing communications and networks, among other features.
Entering in more detail in the Bullx B520 model, some of its elements have been redesigned with the purpose of optimizing both the performance and energy consumption. Meanwhile, and from a technological point of view, some of the advances that have taken place, such as the latest generation (so far) of the Intel Xeon processor family E5-2600 v3, which achieves greater energy efficiency and multiply by three the performance per watt; it allows for faster data frequencies compared to the previous generation using standard DDR4 memories; and provides high-performance computing capabilities for business workloads. The firm is also committed to SSD flash drives and the InfiniBand network connection.
All this approach is going to translate into a series of benefits and advantages, which are indicated below. The first refers to the reduction in the cost of ownership that entails: fewer components, better adjustment of the communication network, faster installation times and improved energy efficiency and updating capacity. Next, we must point out the improvement of the performance, as well as an improved reliability (less cable).
The chassis of this type of solutions can accommodate up to 9 double blades in 7U. It also contains the first level of interconnection and a management unit, among other elements, to build a high-performance medium-large cluster. On the other hand, a performance peak of more than 18 Teraflops per chassis and 110 Teraflops per rack is provided thanks to the family of processors mentioned above, and a bandwidth memory of up to 256 Gb per node.