HPC consulting company X-ISS has released the results of a survey conducted at SC14 about the challenges of operating and managing HPC systems. Participants were asked to rank ten questions on a scale from zero (no problem) to ten (it’s a major issue).
Questions were intended to cover a range of HPC challenges, or “HPC Pain Points” – relating to HPC installations as well as management and support.
In a recent blog post, X-ISS provided a general overview of the survey methodology and statistics, and says that more details will be released in the coming months.
In total, 80 surveys were completed, however only 54 records were completed sufficiently to be usable. The number of nodes on the HPC systems ranged from 2 to 20,000 with the average number of nodes coming out to 1,970. Shared storage ranged from 25 GB – 40 PB.
Total shared storage across all respondents was 144 PB with 21 respondents operating 1 PB or more. InfiniBand fabric is in use by 83 percent of respondents.
The ten questions cover issues related to hardware, workload characteristics, staffing, compliance, security, and more. Questions 1 and 2 were covered in the first results installment:
Question 1 sought input on integration of HPC cluster into enterprise infrastructure. There was a wide variety of results with 22 percent of those surveyed stating no issues with enterprise integration and just over one-third (34 percent) reporting a pain point of 8 or higher.
Question 2 asked for feedback regarding managing multi-vendor hardware, cluster managers and schedulers.
The results on this question were fairly well-dispersed, but notably 52 percent were sufficiently challenged to rate this a 5 or above, with 7 percent scoring it a 10 and 19 percent reporting zero issues. In the opinion of X-ISS, “high diversity of multi-vendor hardware, cluster managers and schedulers often results in larger staffing demands, less standardization, slower support response time and a less performant HPC infrastructure.”
The full results will be reported in the coming weeks at http://www.x-iss.com/blog/hpc-pain-points-survey/.