CMVM Compute and HPC Strategy

CMVM Compute and HPC Strategy vs 1.0

 

Last revised: 04/02/20

 

Document creation sign-off

 

 Name

 Role

Sign off date

Written by:

 Anthony Davie  

 CMVM IS Team Lead, Easter Bush

   23/01/20

 

 David Perry

 CMVM IS Team Lead, Western General   

   04/02/20

Approved by: 

 Sheila Fraser

 Head of IT for CMVM

   04/02/20

 

 

 

 

Cleared for issue by:

 Sheila Fraser

 Head of IT for CMVM

   04/02/20

 

Record of amendments

Version

  Section/s

  Amendment/s

Sign off date

0.1

  All sections

  Creation of strategy, version 0.1  

    23/01/20

1.0

  All sections

  Signed off at CMVM RCG

    04/02/20

 

 

High-Performance Computing (definition)

The practice of combining computing power in a way that delivers much higher performance than achieved from a typical desktop computer or workstation in order to solve complex problems.

 

Vision and mission statement

The College of Medicine and Veterinary Medicine (CMVM) has not had a strategy for compute and high-performance compute (HPC). With compute increasing in demand and the rapid evolution of technology solutions and services, CMVM is at a point where a strategy is needed so that researchers can easily understand and access robust, reliable services.

Where University provisioned research compute services can fulfil CMVM research needs, CMVM will endorse the use of relevant services as the default. The College will assist its research community with a standard approach for the adoption, use and sustainability of such services.

 

Summary

The key steps to achieving the aims of the compute and HPC strategy [1]are:

 

  1. Understand where we are now.
  2. Explore opportunities to work together across CMVM and service providers.
  3. Reduce costs by pooling investment and resource to support future provision of service/s.
  4. Investigate gaps in the existing services.
  5. Achieve and agree a standard approach for HPC across CMVM.

 

Plan

Next steps are based on CMVM research community survey outcomes from Nov ’18 (see Fig 1).

Fig 1: Compute use in CMVM, by compute service type

CMVM compute and HPC survey responses

[1] This strategy directly supports the CMVM Strategic Plan and the CMVM IT Strategy 2019-22.

 

Fig 1 shows that researchers across CMVM use different types of platforms for their compute needs. It is clear from the survey feedback that some of this has been achieved through talented individuals developing local solutions based on their knowledge. In some cases this leads to bespoke solutions and workflows, with similar functions being replicated across CMVM.

Innovation is vital within Research. Whichever platforms are used, it will be important to ensure that System Maintenance, Service Sustainability and relevant Researcher Competencies can be upheld over time.

In addition, where personal computers are used for the full active lifecycle for compute processing; storage and analysis of primary data and results, this increases the risk of compromising quality research outcomes.

 

In order to accomplish the strategic objectives, we will require to tackle some of the issues highlighted in the HPC survey (see Fig 1 above). This can be accomplished by:

  1. Understand where we are now

    1. Review the current Research structure within the college.
    2. Survey Research staff to understand current usage of HPC compute.
    3. Review the issues raised with current UoE services (see survey results document).

 

  1. Explore opportunities to work together across the college and service provider

    1. Working with CMVM IT staff and Computing Officers within CMVM to make sure they:

      1. Understand the HPC services that are available within the UoE.
      2. Can support and troubleshoot basic issues with the usage of the HPC services.
      3. Can install and manage software for their area of CMVM.
      4. Have links to the teams managing and supporting the service.
      5. Build relationships between CMVM IT staff and Computing Officers supporting the service to provide a CMVM support network.

 

    1. Work with the service provider to:

      1. Build relationships with CMVM IT staff and Computing Officers similar to that created through existing CMVM partnerships.
      2. Provide training to Researchers on compute, HPC and Cloud-based compute services.

 

  1. Reduce Total Cost of Ownership by pooling investment opportunities and resource to Support future provision of service

    1. Work with finance to make sure that grants include appropriate charges for the use of HPC, cloud and associated compute related storage.
    2. Where there are large requirements for compute:
      1. Establish requirements and funding.
      2. Where possible/applicable work with service providers to extend or add to partnership agreements.

 

    1. Review existing compute and storage hosted locally, to create a plan for migrating to central HPC, cloud and storage if possible.  Issues or concerns should be raised with service providers to assist if possible.

 

  1. Investigate gaps in the existing services

    1. Review the issues raised in the HPC survey with Computing Officers and service providers.
    2. Liaise with Research Facilitators to work with researchers to try to improve transparency of research computing options and benefits, where appropriate, of using UoE services.
    3. Work with service providers to provide training on HPC and Cloud services for Computing Officers and Staff.
    4. Investigate with service providers, the options for innovation where service gaps exist to tailor existing or create new services to further benefit the CMVM research community.

 

  1. Achieve and agree a standard approach for HPC across college

    1. Recommend the usage of UoE hosted HPC and Cloud services, where these align to delivering requirements.
    2. Work with service providers to make sure any transition to use this is as smooth as possible.

 

Performance Management

In drafting the strategy, it was anticipated that there would be areas of CMVM that were already making good use of the UoE centrally provided HPC services.

To review the success of this strategy, the following three ways can be used to evaluate and measure progress:

  1. Poll for feedback from the CMVM research community via the CMVM Research Computing Group. This could be conducted on an agreed periodic basis.
  2. Conduct further CMVM wide compute and HPC surveys to gather current data, evidence progress and focus further adoption of standards. This should be conducted at 1-2 year intervals.
  3. CMVM IS teams can liaise with compute and HPC service providers to gather service utilisation information, then format this by Institute or Centre to understand point-in-time uptake of services.

 

This strategy should also be reviewed and updated on the same timeline as the compute survey, ensuring it remains current and effective for the CMVM research communities compute and HPC requirements.

AttachmentSize
Image icon CMVM compute and HPC survey responses31.76 KB

Project Info

Project
Compute and High Performance Compute strategy
Code
MVM302
Programme
MVM Research (MVMRES)
Management Office
ISG PMO
Project Manager
Anthony Davie
Project Sponsor
Paul McGuire
Current Stage
Close
Status
Closed
Project Classification
Run
Start Date
08-Aug-2016
Planning Date
n/a
Delivery Date
30-Aug-2019
Close Date
28-Apr-2020
Programme Priority
2
Overall Priority
Normal
Category
Discretionary