TANGO HPC

TANGO HPC
August 2, 2017 Lauren Attana

The new TANGO High Performance Computing (HPC) is part of a specialised hardware cluster built using a combination of Dell EMC and VMWare, hardware and software technologies to deliver a powerful computational
solution.

TANGO HPC

This platform consists of core technologies that allows us easy and fast expansion for the delivery of high performance computational resources.

It has the added advantage of latest generation Intel cores specifically designed for parallel processing with a high core to memory ratio, above that of traditional HPC environments.

Key benefits

  • Single high performance software defined network fabric which facilitates easy scalability for high performance data centre and compute environments.
  • The new Intel Xeon processors E5 v4 product family has had vast improvements to the vector floating-point multiply and scalar divide. It can deliver up to 47% more performance across a wide range of HPC codes based on Intel’s internal testing, without the need to recompile existing code.
  • High core to memory ratio for standard HPC CPU workloads, allowing higher memory jobs to utilise the standard HPC queue, without the need to use Big Memory nodes which are in high demand.

Technical specifications

TANGO HPC cluster is built utilising:

  • Dell PE R730 Servers connected via Dell Z9100-ON 100GbE Software Defined Network switches
  • Purpose-built for applications in high-performance data centre and computing environments.

     The easily scalable cluster currently consists of:

  • Intel Xeon E5 cores
  • 1:9 Core to RAM ratio
  • The compute nodes each have 400GB of local SSD storage
  • Connected to VMware VSAN node cluster via NFS Virtual Machine gateways hosting storage for scratch and cloud, with each node being capable of 20,000 IOPS
  • Message Passing Interface (MPI) for parallel computing jobs has been implemented using RDMA over Converged Ethernet (RoCE) removing the need to run a dedicated InfiniBand storage network utilising the Dell 100GbE backbone
  • Big Memory nodes ranging from 256GB to 1TB

System specifications

Hardware Dell PE R730
Specification 

  • Intel® Xeon® Processor E5-2690 v4 35M Cache 2.60 GHz
  • 32GB RDIMM 2400MT/s DDR4 SDRAM
  • 200GB SSD 6Gbps
  • Mellanox ConnectX-4 Dual Port 25GbE DA/SFP

  • “The supercomputing facilities at eResearch SA permit analysis of a host of interesting problems in evolutionary biology. It is the only computer system in SA that can perform certain complex calculations required to infer large evolutionary trees and associated patterns of evolution.”  
    Associate Professor Michael LeeSouth Australian Museum
  • "eResearch capabilities ... ensure we can continue to use the latest methods available in our field. The hope is that through the use of these technologies, we will be able to achieve some research outcomes that may otherwise not have been possible.”  
    Professor Ina Bornkessel-SchlesewskyCognitive Neuroscience, UniSA
  • "The eRSA support is very personal and solution-focused and not just a brushing off 'it's-your-fault-check-your-code' kind of support which one sometimes gets from University tech supports. I very much appreciate the help."  
    Sven SchellenbergSchool of Science, RMIT University
  • “It would be impossible to do the type of research that we’re doing without them – it is a major factor in achieving our research outcomes.”  
    Associate Professor Con DoolanSchool of Mechanical Engineering, University of Adelaide
  • On using an eResearch program to complete 3D modelling of architectural records: “We were able to build an interactive, photo image model of the [historical South Australian property Joseph Elliot’s cottage] home – transforming simple sketches and floor plans into an interactive 3D experience... what we have now is a computerised model that allows us to experience the Elliott house beyond just words on paper”.
    Associate Professor Christine Garnaut
  • “Open source intelligence is about extracting information from blogs, news sites and social media platforms, any information that can be freely accessed online. There’s way too much information out there for an individual or group of people to comprehend, so we have created automated tools to allow our users to extract the data they need...access to the latest technology allow[s] us to continue to provide the best platforms to our end-users.”
    David BlockowData to Decisions CRC

Our partners

University of South Australia logo