HPC-SIG Meeting, 2023-03-07, Southampton

Posted on by Andrew Turner

Location: University of Southampton

HPC-SIG Southampton March 2023

Welcome and Chair’s Business 

  • Ed briefly reviewed the recent SIG meetings (Birmingham and online Q&A). June 2022 Birmingham meeting, Nov 2022 online QA, CIUK sysadmin/infrastructure engineer session, March 2023 HPC-SIG coordinated online HPC/DRI meeting 
  • Ricky gave a financial update: 
    • Subscription fees to remain at £150 
    • Agreed to waive fees for the first two years for new institutions 
    • Current balance stands at £41,449.94 
  • Upcoming costs potentially £5K for potential mentoring scheme 
  • Potential upcoming costs in engaging a consultant to advise on external engagement and branding. 
  • Steven set out the idea for non-exec committee posts and a new exec EDIA role. There was then a vote on the relevant changes to the terms of references, which passed unanimously. 
  • Changes approved to changes to terms of reference on the appointment of non-exec committee posts (e.g. roles to help with website, roles to organise one off events, role from host institution of next SIG meeting) 
  • Approval also of addition to Terms of Reference to include a new Equality, Diversity, Inclusion and Accessibility (EDIA) Officer. The committee will now look at the election process for this role ahead of the next meeting. 

Reframing Diversity, Equity, and Inclusion in HPC: Cristin Merritt (WHPC, Alces) 

  • Invited speaker. Cristin shared her views on sustainability in the technical and HPC fields. 

Southampton Presentations  

Slide deck: 2023-03-07_HPCSIG_Soton

  • Exploring how they can use SLURM to address challenges with sustainability and sharing their experiences.  
  • Understanding where power is being consumed. If nodes are idle then turn off. Discussed how this has impacted on reliability of their systems, by turning nodes off and on.    

Open OnDemand: Thomas Green and Simon Branford  

  • Slides will be shared soon

Parallel Session: Technical 

Lightning talks  

Slide deck: 2023-03-07_HPCSIG_Lightning-Technical

  • “Modular/Containerised Data Centre (The Solution/The Future?)” Simon Atack, HPC Team Leader, ACRC, Bristol University 
  • “Who Needs Infiniband? A (slightly) tongue in cheek argument for converged Ethernet only HPC !”, Jonathan Churchill, SCD Network Architect, Scientific Computing Department, STFC – Rutherford Appleton Labs. UK 
  • “Virtual Environments – Teaching to Fish” Taylor Haynes, Research Computing Systems Engineer, University of Southampton – HPC – iSolutions 
  • “HPC at the Norwich Science Festival”, Jake Newman, Research and Specialist Computing, University of East Anglia 
  • “ARCHER2: Application efficiency monitoring with ReFrame”, James Richings, EPCC, The University of Edinburgh  

Future topics  

  • Kubernetes, K8s – Requested with offers to cover in the future from Thomas Green (Cardiff) and Matt Singh (Kingston) 
  • Jobsubmit.lua – Requested 
  • Spank plugins – Requested, with offer from Sam Gallop (UEA) 
  • Slurm compatible task profiler (CPU, memory, GPU) – Volunteered by Sam Gallop (UEA)  
  • Capacity Management (technical perspective) – Requested, with suggestion Andy Turner (EPCC – absent) may cover 
  • Trusted Research Environments – Requested  
  • Research Data Management – Requested  
  • Software Lifecycle Management – Volunteered by James Cross (UEA – on behalf of colleagues) 
  • CPU clockspeed/power utilisation update in production (based on Application efficiency lightning talk) – Volunteered by James Richings (EPCC) 

Parallel Session: Management/strategic 

Lightning talks 

  • “RUGRIT & HPC-SIG – Service Provision ‘Sustainability’ Survey Sep. 2021, A Lightning Summary”, Andrew Edmondson, HPC-SIG Chair 
  • “Help with building a new HPC community in Southampton”, Simon Hettrick, University of Southampton 
  • “Building Blended Research Computing Services – Supporting and enabling research activities” Dr Andrew Richards, Head of Research Computing Services, Imperial College London 

Discussion  

  • Discussion largely revolved around experiences with engaging with the academic communities and influencing senior leaders. What have people found which works, such as being able to link ideas/strategy with a story or a narrative. Some sites have found producing an annual report very useful. It was noted that we should lean on RSEs who have good case studies to write up.  

Future topics  

  • Engagement session. Effective ways to engage with the academic/research communities.  
  • Cyber Essentials  
  • How are we sharing training material and best practices within the SIG? 
  • Career development, succession planning session – potential volunteer Owain Kenway (UCL) on their experiences. 
  •  

CIUK sysadmin breakout sessions  

  • A review of the two sessions at CIUK22 organised by Simon Atack was presented by Simon and Chris Collins 
  • The sessions proved to be very popular with approximately 50 attendees on day 1 and 20-30 on day 2 
  • No set agenda, just discussion on topics of interest that arose – Slurm power monitoring and powering off, HPC build tools (Spack and EasyBuild), the HPCSIG community 
  • Support for a similar, more formal (but same style) event at CIUK23, possibly getting on the program maybe a breakfast style event. 
  • There was also a suggestion of a possibly regional BoF at ISC24 
  • A quick poll around the SIG meeting attendance indicated around 20 had attended CIUK with around 10 attended the sysadmin get together. A further 15 people suggested a colleague should attend such a session next time. 

Wrap up, notices  

  • We noted the upcoming hybrid HPC-SIG meeting in Bath, on 13th June. 
  • Request for volunteers who can host meeting, with particular interest in those with good hybrid capabilities.