High-performance computing clusters are computer networks that can perform a large number of computing operations in parallel. At the Museum für Naturkunde Berlin, eleven computers with a total of 567 processor cores and 4661 gigabytes of RAM are linked with two servers. The servers are used for management purposes. They can be used to manage users and distribute computing resources efficiently and fairly among the research groups.
The combination of the devices allows different requirements to be covered. This means that time-consuming calculations that can take more than 100 days can run in parallel with calculations that are distributed across several processors or require a lot of memory.
The use of the cluster can be easily integrated into the researcher's workflows: Access and data exchange are possible from any computer workstation in the museum. The cluster's operating system and other software such as editors, tools and program libraries enable a wide range of scientific applications.
Methodology (selection):
- Numerical modelling of highly dynamic processes
- DNA sequence analysis
Equipment:
Computing power of the cluster
Nodes: 6x Dell PowerEdge R815 with each:
4 x AMD Opteron(tm) Processor 6172 with a total of 48 CPU cores, 256GB RAM (DDR3-1333)
1x Dell PowerEdge R815:
4 x AMD Opteron(tm) Processor 6278 with a total of 64 CPU cores, 256GB RAM (DDR3-1333)
1x Dell PowerEdge R615:
2x Intel(R) Xeon(R) CPU E5-2690 with a total of 16 CPU cores and 32 threads, 384GB RAM (DDR3-1600)
1x Dell PowerEdge R7425:
2x AMD EPYC™ 7501 CPU with a total of 64 CPU cores and 128 threads, 768GB RAM (DDR4-2666)
Virtual nodes: 2x2 Intel® Xeon® Processor E5-4667 v4 with a total of 64 CPU cores, 885GB RAM (DDR4)
Management: 2x server for cluster management:
1x Dell PowerEdge R310:
1x Intel(R) Xeon(R) CPU X3450 with 4 CPU cores and 8 threads
16GB main memory (DDR3-1066)
1x Dell PowerEdge R310:
1x Intel(R) Xeon(R) CPU X3450 with 4 CPU cores and 8 threads
32GB main memory (DDR3-1066)
Applications:
- The high-performance computing cluster made it possible to reproduce experimental investigations in numerical models within the MEMIN project (Multidisciplinary Experimental and Modelling Impact Research Network). By combining geological and geophysical observations with experimental investigations and numerical models, a comprehensive understanding of the dynamic processes during an impact event was obtained. The models provide detailed information on shock wave propagation as well as temperature and pressure distributions in the rock, which cannot be investigated dynamically in the experiment. Furthermore, so-called mesoscale models were used for a detailed analysis of shock wave induced pore collapse and the investigation of material behaviour. Results of mesoscale modelling were used as benchmarks to calibrate macroscale models for sandstone.
- The cluster was used for the GENART project (Functional Genomics of Biological Speciation) project for various bioinformatics, genomic analyses. It was shown that metabolic processes play an important role in the ecological differentiation of locust populations. New genome regions could also be found in crickets, which influence species-specific songs and probably contributed to speciation. This is an example of evolutionary biological applications in which not only genetic markers are analysed, but entire genomes are examined for variability. This involves very large amounts of data that can only be processed with high-performance computers.
- The Collaborative Research Centre (SFB) Transregio 170 “Late Accretion onto Terrestrial Planets” deals with geodynamic and geochemical processes in the early history of terrestrial planets. The thermal evolution of the earth-moon system and especially the melting during impacts are investigated. With the iSALE software, which has been developed and hosted at the Museum für Naturkunde for more than ten years, several hundred impacts of asteroids, ten to thousands of kilometres in diameter, were simulated to determine the mass of molten and vaporized material depending on various initial thermal conditions. Results indicate that the mass of impact-generated melt and steam can be up to seven times greater than previously estimated. Further simulations are needed to track the whereabouts of the ferrous impact core. On the basis of the findings, the impact history and early development of the Earth can be reconstructed from today's geochemical fingerprint of the Earth's mantle.