DATA INTENSIVE SERVICE
(LEICESTER)

The Data Intensive Service offered by DiRAC encompasses a range of hardware options tailored to projects spanning all our research domains. Its primary aim is to facilitate scientific discovery by significantly enhancing our capacity to manage and process large datasets. This includes conducting and analysing precise theoretical simulations and comparing them with the latest observational and experimental data. To effectively support such diverse workflows, the Data Intensive Service employs a heterogeneous mix of computing architectures, which are distributed across two DiRAC sites: Cambridge and Leicester. 

System name: DIaL

In addition to standard nodes with up to 512 GB of RAM, the Leicester service also provides access to a set of nodes with significantly higher memory (up to a maximum of 6TB). All nodes access the same tightly-coupled parallel file system, allowing workflows to use both regular nodes and higher-memory nodes seamlessly. 

Leicester also led DiRAC’s exploration of alternatives to x86-based CPU processor technologies. Hosting an HPE/Arm/Suse Linux catalyst Arm-based clusters permitting  testing of DiRAC workflows on this technology. 

The benchmark codes used for the design and testing of the DIaL system were: 

TROVE – a state of the art code for calculating molecular line lists, requiring the diagonalisation of enormous O(106x106) matrices. Workflows exploit the large memory nodes. Results are used to interpret spectra from extrasolar planets, in particular to look for signs of water or life. 

RAMSES – a large-scale cosmological simulation code, used to study the complex processes driving galaxy evolution, including the growth of supermassive black holes and the role of magnetic fields.  

SphNG – a code to simulate star formation as realistically as possible.