Computational Resources

The NUS Centre for BioImaging Sciences HPC computational resources are designed to further research activities that include: collection and storage of very large datasets (TeraBytes) generated from optical and electron microscopy imaging systems, analysis and modeling of the data using complex computational methods, and interactive visualization of the results.

CBIS2

Large imaging data sets present significant challenges for traditional computational environments to efficiently manage. Thus we’ve designed a computational infrastructure optimized to store, archive, process and transfer very large datasets, enable fast user interactive access and provide a user friendly experience.

The facility is housed in a modern 50 sq m  server room the provides the needed power, cooling and monitoring systems to support next generation high-density compute and storage servers.

Current CBIS computational resources

Currently CBIS’ computational resources include two multi-processor, large shared memory computational servers from SGI for the analysis and modeling of the data:

  • 96 core, with 1 TB mem, providing 2.0 SP TFLOPS
  • 80 core, 2 Phi co-processors with 1.28 TB mem, providing 7.3 SP TFLOPS
ITcore-baiChang
Bai Chang in the CBIS server room.

– ~500 TB high performance DDN storage arrays from SGI (expandable to PetaBytes) for the collection and archival of optical and electron microscope, and Next Generation Sequencer datasets. Storage arrays are directly connected to high-end compute and storage servers through a dedicated 8 Gb/s Fibre Channel Storage Area Network.  The storage assets are arranged in multiple tiers with the location of files within those tiers under control of a Hierarchical Storage Management system ( SGI – DMF), which automatically migrates data files between the stores based on a set of rules and user demand.

– A 6 Tb/s Ethernet core switch provides the Centres network backbone with multiple 10 Gb/s connections to the NUS campus network, CBIS servers, CBIS microscope facilities, other DBS core facilities and CBIS edge switches that provide dedicated 1 Gbps links for each user.

– Mediaflux digital archive management software (Arcitecta) enhanced with the OME data interface (SGI) for the storage and management of our microscopy data sets.

The CBIS data centre is connected to the Mechanobiology Institute (MBI) data centre through the NUS campus backbone. This connection utilizes multiple high-speed 10 Gbps Ethernet circuits allowing us to share the computational and stoarage resources of the two facilities.  This capability allows users the freedom to run experiments and collect data at either research Centre and process their data without using local compute resources, minimizing the need to move very large data sets across the network.