Skip to content

Statistics of the HPC

Notes and Terminology

CPU Hour

A CPU Hour is a scale determining how long a single CPU will be utilised at 100% to perform a specific calculation. Thus, a CPU Day indicates how many physical days a calculation would run on a single CPU Thread/Core.

node

A node in an HPC is simply a large server that usually has several gigabytes (or terabytes) of System Memory (RAM) and several CPUs per server.

The data displayed on this page only reflects the CPU hours and no GPU Hours.

Summary

Description Value
Number of Jobs 4287
CPU Hours for 2025 91y 298d 3:27
Users (active / total) 112 / 807
Available Software Versions / Packages 147 / 566
Available CPU Cores 5544
CPU Hours since 2010 9 558y 320d 4:17

The above table reflects the system usage statistics for the current year.

Top Ten Users

User Jobs Process Hours Average Hours per Job
#207 Physics 27 302 742.5 14 343.38
#330 Physics 369 226 242.8 613.12
#114 Chemistry 434 97 833.4 225.42
#307 Chemistry 111 51 587.0 5 405.58
#248 Medical Physics 2186 24 619.9 11.26
#605 Student 102 13 663.2 134.32
#593 Office Of The Dean: Natural Sciences 5 10 735.5 2 147.10
#187 Mathematical Statistics And Actuarial Science 53 8 827.1 198.52
#377 Medical Physics 378 8 632.3 75.79
#633 Master Of Medical Science With Specialisation In Medi 30 4 164.0 146.48
User Jobs Process Hours Average Hours per Job
#248 Medical Physics 2186 24 619.9 11.26
#114 Chemistry 434 97 833.4 225.42
#377 Medical Physics 378 8 632.3 75.79
#330 Physics 369 226 242.8 613.12
#307 Chemistry 111 51 587.0 5 405.58
#605 Student 102 13 663.2 134.32
#142 HPC 62 3 484.2 56.20
#187 Mathematical Statistics And Actuarial Science 53 8 827.1 198.52
#600 Genetics 37 2 112.6 64.01
#974 Genetics 35 1 075.8 41.46
User Jobs Process Hours Average Hours per Job
#207 Physics 27 302 742.5 14 343.38
#307 Chemistry 111 51 587.0 5 405.58
#593 Office Of The Dean: Natural Sciences 5 10 735.5 2 147.10
#982 Medical Physics 8 3 972.2 792.18
#330 Physics 369 226 242.8 613.12
#372 Chemistry 1 580.2 580.20
#101 HPC 2 747.7 534.50
#342 External 7 3 624.9 517.84
#268 Medical Physics 3 1 200.3 400.10
#849 Information And Communication Technology Services 9 2 261.8 353.80

Above, three statistic measurement groupings indicate the utilisation of the HPC per (anonymised) user.

The groupings are as follows:

  • CPU Hours
    • This is the total usage per user.
  • Number of Jobs
    • This value indicates which users are submitting the most significant number of jobs. In an HPC, monitoring the number of job submissions is essential to ensure that a fair share of resources is in place.
  • Average CPU Hours per Job
    • This value is essential to ensure that users (submitting thousands of smaller jobs) do not exhaust the available resources for larger jobs. It also indicates which users run larger simulations that can take up to several months, utilising several nodes in parallel, during which the resources are unavailable to other users.

CPU Hours for 2024/2025

{ "$schema": "https://vega.github.io/schema/vega-lite/v6.json", "data": { "url": "data/stats_cpu_hours.json" }, "width": {"step": 40, "for": "position"}, "mark": "bar", "encoding": { "x": {"field": "Department", "axis": {"labelAngle": 300} }, "y": {"field": "Value", "type": "quantitative", "sort": {"op": "sum", "field": "Value", "order": "descending"} }, "xOffset": {"field": "Year"}, "color": {"field": "Year"} } }

The above graph shows the utilisation of the HPC per department for the previous year two years. When a significant decline compared to the prior year is observed, it usually indicates one or two PhD candidates in the specific department finishing their PhD during the last year.
Activity from training accounts is often observed in the table. However, these accounts are not purely training accounts. Specific departments with complex software requirements sometimes use them. Assigning the workflow to a named account holder may be too time-consuming or too complex for the purpose, or sharing datasets between several users performing a specific function may entail a storage burden that is too significant.

Note that the previous year's values are the total for that year and are not compared for the same period for that year. So, a value up to June indicating that there is about half the amount of processing performed on the HPC vs the previous year suggests that the utilisation is like the prior year and that the total at the end of the year will be near the value of the previous year.

CPU Hours Per Department (Since 2010)

{ "$schema": "https://vega.github.io/schema/vega-lite/v6.json", "description": "Statistics per department since 2010.", "data": { "url": "data/stats_since_2010.json" }, "mark": "bar", "encoding": { "x": {"field": "Department", "type": "nominal", "axis": {"labelAngle": 300}}, "y": {"field": "Value", "type": "quantitative", "sort": "-x" } } }

The above graph indicates all the CPU hours performed per department since 2010. This graph shows that Chemistry, Physics, Animal-And-Wildlife Sciences, and Medical Physics are the largest consumers of HPC CPU resources.