High Performance Computing Glossary - H


Haswell is an Intel processor micro architecture, based on 22nm die shrink. Haswell is the predecessor of Broadwell.

Heat Transfer

Heat transfer has broad application to the functioning of numerous devices and systems. Heat-transfer principles may be used to preserve, increase, or decrease temperature in a wide variety of circumstances. Heat transfer methods are used in numerous disciplines, such as automotive engineering, thermal management of electronic devices and systems, climate control, insulation, materials processing, and power station engineering.

High Performance Computing (HPC)

High-performance computing (HPC) is the use of parallel processing for running advanced application programs efficiently and quickly. The term applies especially to systems that function above a teraflop or 1012 floating-point operations per second.

HPC becomes more important in the fields of commercial and scientific computing as a means of calculation, modelling and simulation of complex systems and of processing big amounts of data. Typical applications can be found in CAE, e.g. crash tests or CFD, as well as in meteorology, astrophysics, biology, quantum chemistry and genetics.

HPC Cluster

HPC clusters in a CPU 24/7 high security computing centre usually consist of one or more head nodes and several compute nodes, which are freely scalable according to the application and specific requirements.

The head node is a server in the HPC cluster, on which the individual calculations are distributed to the compute nodes (link) with the aid of software. Head nodes may also be referred to as access servers, because they are usually used for remote visualisation.

The compute nodes are servers in the HPC cluster with high performance CPUs, on which the computations or jobs are carried out. They are also equipped with a large memory and are connected to other nodes by means of a high performance InfiniBand network.

The head and compute nodes are preceded by one or more security access gateways and adaptive firewall appliances, such as VPN. A high performance, parallel file system with data tiering, several replication levels and a native InfiniBand connection form the basis for the HPC cluster. The HPC cluster is completed by management units, consisting of a choice of a licence server, GPU server or queuing system, as well as many other items.

Please note that CPU 24/7 HPC clusters are not based on a shared IT infrastructure and consequently their use is confined to the specific client. Customers and their data will be isolated by completely separate systems. Do you require other individual components? No problem! We will provide the basic framework and you supply us with the details.

HPC On Demand

HPC On Demand is a means for companies to temporarily access scalable HPC resources that can be activated and intensively used as burst capacities for CAE projects. For users the acquisition of comparatively cost-intensive hardware and specific software solutions represents a large investment, tying up capital, expertise and manpower in areas outside those of the company’s main business activities.

CPU 24/7 offers many solutions for meeting HPC On Demand requirements, all of which are simple, secure, readily available and perfectly designed to meet your requirements. Remote and on-demand, for engineering consultancies and major clients alike.