Initial results of the Scientific Computing World (SCW) survey for HPC resource managers have shown that seeking investment and training and support are major concerns for those that operate HPC facilities.
The survey has now been live for around two weeks and early results are helping to highlight topics which represent the challenges that managers face in supporting HPC research.
Several themes have emerged as a priority to those running HPC facilities. Respondents cite that ‘seeking investment’ and ‘training and support’ are the most prominent challenges. However, software issues, cluster management and demonstrating ROI were all cited as primary concerns for our readers who are operating HPC resources for research.
Respondents were asked to highlight other issues they face in the survey and several patterns have emerged in different areas which also highlight the challenges faced by the teams running HPC resources. One consistent message was managing expectations of researchers and providing feedback to end-users on cluster usage and metrics on individual jobs, or increasing user experience.
Several users also reported that going beyond the standard batch environment or command-line interface to provide a more service orientated model - particularly for new to and inexperienced users is a significant challenge.
Management of the software stack, integrating workflows, code optimisation and code performance, software quality and data management were some of the more traditional issues associated with running HPC facilities.
There were also several comments tied to investment or demonstrating ROI for HPC systems. Potentially exacerbated by the current viral pandemic, respondents report trying to meet research needs with limited staff, needs to reduce operational costs and having to deal with end-of-life hardware as additional challenges.
More than 94 per cent of respondents report that they are using an in-house cluster. While this is not 100 per cent representative of the entire HPC community it highlights that while cloud computing technologies are pervasive they are primarily used to enhance and not replace existing HPC capabilities.
More than 60 per cent of respondents cite that seeking investment is either very challenging or a ‘major headache’ compared to 44 per cent who report software issues as their primary challenge. Management software, training new and existing users and demonstrating ROI were reported as either very challenging or a ‘major headache’ by 29,28 and 27 per cent of survey participants.
The survey is still available and SCW encourages anyone who operates or manages HPC resources to take part as this will help us to create content that more accurately reflects our readership.
