SERVICES

Summary

High-Performance Computing (HPC) is a fundamental basis for many scientific disciplines and shows growing demand. In Germany, a multi-layered supply has been defined to meet this demand, consisting of Tier-1 computers for cutting-edge research and a national high-performance computer structure (NHR) as Tier-2 for large projects on an application basis. Another layer, Tier-3, includes local systems, e.g. at universities, as well as state infrastructure for general HPC requirements. Supply in Lower Saxony has so far been inadequate at this level. Furthermore, there is currently no overall concept to enable researchers to ensure effective and efficient handling of parallel computers on a broad front. These gaps are being closed with the HPC.NDS project.

Through the HPC.NDS project as part of Hochschule.digital Niedersachsen (HdN), high-performance computing systems (HPC) in the form of hardware and/or consulting services through so-called Science Support Teams are made available at a total of six locations in Lower Saxony:

The operation of the computing resources is coordinated by GWDG in Göttingen and the consulting network Lower Saxony Digital Science Support Space (DS³) by TU Braunschweig. HPC.NDS offers the following advantages for the target groups, users and universities in Lower Saxony.

1 HPC Users in Lower Saxony

Advantages of HPC.NDS

For cutting-edge research and teaching, university members need a low-threshold offer that provides uncomplicated access to powerful basic HPC infrastructure with guaranteed availability. They also need development and application software, consulting and training on the use of the systems, and efficient support for technical problems.

Therefore, HPC.NDS not only provides computing resources, especially CPU and GPU nodes, but also offers comprehensive support and within the framework of DS³ various consulting services with a focus on Research Software Engineering (RSE) for HPC-suitable scientific software. Beginners thus have easy HPC access, while already productive workloads benefit from the additional capacity provided through this project.

What advantages does this form of HPC access offer?

The HPC.NDS resources represent a supplement to possibly existing local HPC resources (Tier-3). Unlike the Tier-1 and Tier-2 systems, no complex application is required for access and the systems are just as easy to use as local computing resources. There is an uncomplicated booking procedure for resources at the operating centers. If needed, contact your university to book and use HPC.NDS capacities. Hardware procurement in HPC.NDS is based on actual demand; appropriate feedback can be given to the operating centers at any time.

2 Universities

Advantages of HPC.NDS

The demand for HPC resources, whose commitment is often part of appointment promises, is increasing at universities. These additional resources must either be procured and operated by the local computer center or by the institutes themselves, which requires corresponding investment funds for hardware and infrastructure and operating funds for personnel and energy. With HPC.NDS, the situation is simplified: only the energy costs still have to be borne by the participating universities. In addition, shares in HPC.NDS can be exchanged with other participants.

The Tier-3 HPC resources of the HPC.NDS operator centers flexibly supplement any existing local Tier-3 resources. Procurement is organized according to needs, ensuring an appropriate selection of node types (memory equipment, accelerators, etc.). Operation is ensured by the operator centers; it includes the availability of IT infrastructure, installation and updates of systems, provision of system software as well as current and optimized development tools and application software.

Under what conditions is participation possible?

First, an interested university requests a share of the HPC.NDS capacity for use by local users at one of the operator centers. This can then be divided between the operator centers to depict technical requirements of users known in advance (e.g. GPU generation, memory equipment or special accelerator hardware) on the one hand and to ensure sensible load distribution between the computer centers on the other.

While the investment and personnel costs are borne by the HPC.NDS project, the reservation of a capacity share involves the assumption of the associated operating costs by the participating university for an initial period of five years. Billing is initially done via a monthly lump sum, at the end of the year a final invoice is made based on actual consumption. The lump sum is conservatively estimated and is based on 75% of the theoretically maximum possible consumption, thereby minimizing the risk of additional payments. To better estimate costs during ongoing operation, consumption information (actual use and costs incurred based on current electricity price) is reported monthly to react with a usage adjustment if necessary and thus implement cost control.