Center for Research Computing
Research Computing Core Rates: vHPC
The rate to use for calculating the cost of using the vHPC is dependent upon the owner of the account. Rates are increased for academic projects outside of the USG and for other external projects. Total costs are reported in Research Computing Units (RCUs) where 1 RCU = 1 cpu-hour = $0.03.
| Service | Unit | KSU Rate | Other Academic | Other External |
|---|---|---|---|---|
| vHPC CPU | 1 core-hour | $0.03 | $0.04 | $0.05 |
| vHPC GPU | 1 GPU-hour | $0.12 | $0.17 | $0.22 |
| vHPC Mem | GB/job | $0.004 | $0.006 | $0.007 |
vHPC Billing Examples
On the first of each month, completed job information is collected from the previous month. Details about the CPU, GPU and memory resources dedicated to each completed job are used to determine the cost in dollars and RCUs (1 RCU = 1 core-hour = $0.03)
5 Jobs from a KSU-based account
| Job # | CPUs | GPUs | Memory | Duration |
|---|---|---|---|---|
| 1 | 1 | 0 | none* | 1 hour |
| 2 | 1 | 1 | 16 GB | 1 hour |
| 3 | 4 | 2 | 480 GB | 1 hour |
| 4 | 48 | 4 | 256 GB | 1 hour |
| 5 | 48 | 0 | 360 GB | 1 hour |
*If not supplied in job request script, the default memory is set to 5 GB per node,
regardless of the number of cores requested.
Costs for the 5 Example Jobs
| Job | Costs | Total Cost | Notes |
|---|---|---|---|
| 1 |
CPU = 0.03 |
$0.03 or 1 RCU |
CPU + GPU cost greater than Mem cost |
| 2 | CPU = 0.03 GPU = 0.12 Mem = 0.064 |
$0.15 or 5 RCUs |
CPU + GPU cost greater than Mem cost |
| 3 | CPU = 0.12 GPU = 0.24 Mem = 1.92 |
$1.92 or 64 RCUs |
Mem cost greater than CPU + GPU cost |
| 4 | CPU = 1.44 GPU = 0.48 Mem = 1.02 |
$1.92 or 64 RCUs |
CPU + GPU cost greater than Mem cost |
| 5 | CPU = 1.44 GPU = 0.00 Mem = 1.44 |
*$1.92 or 64 RCUs |
*Even though CPU + GPU = Mem, job charged for making 4 GPUs unavailable. |
*The billing process takes into account what resources were made unavailable to other users by a job. So, Job 5 is actually charged for 48 CPU and 4 GPU.
Formula = max(GPUs * GPU rate + CPUs * CPU rate, Memory * Memory rate) * time
If you have long running jobs and would like to know the current cost of the job or are interested in optimizing resources for a job, please reach out to tboyle@kennesaw.edu by email or on Teams.