CaStLeS

What to include in your application

CaStLeS resources (both compute and storage) are a constituent part of BEAR Cloud and the wider BEAR infrastructure. They are reserved exclusively for the use of research groups carrying out research in the life sciences and governed by academics through the CaStLeS Executive and Strategic Oversight Group.

To apply for resources, a project PI should log into the Service Desk at http://www.itservicedesk.bham.ac.uk and under ‘Make a Request’, go to the Advanced Research Computing service and select ‘BEAR/CaStLeS Projects’. This form should be completed, outlining the projects needs and the academic case. Once completed, the application will be forwarded automatically to the CaStLeS Exec for approval.

As a guide, typically, an application will include the following

Details of the Research Group – College, School/institute/collaborators

  1. A short summary of the academic case – the nature of the research, the funder, goals and expectations, timespan of the project
  2. Compute resource required
    a) An explanation of the compute/data intensive nature of the work
    b) Virtual Machines: Refer to the Standard VM page for details of the current cloud offerings.
    c) BlueBEAR CaStLeS QOS: CaStLeS have purchased significant dedicated resources in BlueBEAR for powerful batch computing. Refer to the CaStLeS QOS page for details.
  3. Storage resource required – how many TBs and whether data must be replicated between data centres (to maximize availability) or single sited (cheaper but with more susceptible to interruptions). If you are dealing with sensitive data that has specific requirements e.g. for encryption, make this clear. Unless you specify, the default assumption is that data holdings are NOT sensitive
  4. Applications required
  5. Identify the individuals who are to have access to the resource (UoB usernames are required for validation)
  6. Timescales - CaStLeS resources are in place so, once approved, can be set up with only a short lead-time. However, note that research groups are expected to include the costs in funding bids and contribute accordingly.
Before the resources are provisioned, the CaStLeS Exec must approve the resource request and Advanced Research Computing will complete a technical review to pick up any technical obstacles or queries. Once in service, utilization will be monitored and adjustments made to ensure the facility is operating efficiently.

Brief Summary of CaStLeS Resources – 
potentially useful for research funding applications

CaStLeS consists of 10 compute nodes, each with 128GB RAM, 2x E5-2640v4 (10 core 2.4GHz CPUs) as well as 2 nodes with 1TB RAM, 2x E5-2640v4 (10 core 2.4GHz CPUs) one of which also includes 2 NVIDIA Tesla P100 GPUs 16GB Pascal CUDA PCIe for accelerated Deep Learning. Soon to be released is one Power9 PowerAI System with 1TB RAM and 2 x 18 core 2.6 GHz CPUs and 4 Nvidia Tesla V100 GPUs 

This compute environment accesses 3PB of shared storage, 1PB of which is synchronously copied to a second data centre for maximum availability and all of which is redundantly backed up.

 


Last modified: 1 October 2018