On the pages in this category you will find all the information about ZIMT’s scientific computing resources other than the OMNI cluster. These are the following:
Some resources are integrated into the OMNI cluster to some extent (for example by mounting the same home directory). The similarities and differences are described on the corresponding subpage.
If you would like to know whether using this system makes sense for you, contact us. We will gladly advise you.
Here is a short overview:
The term High-Throughput Computing (HTC) means the computation of a large number of small compute jobs which are usually independent of each other (trivially parallelizable). This is in contrast to High-Performance Computing (HPC) which usually means a small number of large jobs, with interconnected subtasks.
If you have a lot of small jobs with little interaction (e.g. without MPI communication), the HTC system might be attractive for you.
ZIMT operates an HTC system of type HPE Moonshot. Like all compute resources of ZIMT it is available for all university members. It consists of 45 nodes, of which 4 are login nodes and the others are compute nodes. Each node has 8 coers and 64 GB of RAM.
You can find advice on using the HTC system here.
A vector processor is a processor which can perform specific compute operations for multiple numbers at the same time (in a vectorized fashion). Vector processors can perform fewer kinds of operations, but those can be done considerably faster. Typically a vector system has a normal CPU and one or more vector processors. GPUs are an example of a commonly used vector processor type.
ZIMT operates a vector system of type NEC SX Aurora Tsubasa. The system contains two vector cards (called vector engines by the manufacturer).
You can find advice on using the vector system here.