The MOC offers you a pay-as-you-go approach for pricing for our cloud infrastructure offerings (Tiers of Service), Platform-as-a-Service (PaaS) – Red Hat OpenShift. The exception is the Storage quotas on MOC Storage Tiers, where the cost is determined by your requested and approved allocation values to reserve storage from the total storage pool. The MOC offers a flexible cost model where an institution (with a per-project breakdown) is billed for the duration of the specific services requested. Access is based on project-approved resource quotas, eliminating runaway usage and charges. There are no obligations of long-term contracts or complicated licensing agreements. 

If your project is part of the Open Accelerator Program, please click here.

Calculations

Service Units (SUs)

Breakdown

CPU/GPU SUs

Service Units (SUs) can only be purchased as a whole unit. We will charge for Pods (summed up by Project) on a per-hour basis for any portion of an hour they are used, and any VM “flavor”/Pod reservation is charged as a multiplier of the base SU for the maximum resource they reserve.

GPU SU Example:

• A Project with:1 A100 GPU, 24 vCPUs, 95MiB RAM, 199.2hrs
• Will be charged:1 A100 GPU SUs x 200hrs (199.2 rounded up) x $1.803$360.60

OpenShift CPU SU Example:

• Project with 3 Pods with:i. 1 vCPU, 3 GiB RAM, 720hrs (24hr*30days)ii. 0.1 vCPU, 8 GiB RAM, 720hrs (24hr*30days)iii. 2 vCPU, 4 GiB RAM, 720hrs (24hr*30days)
• Project Will be charged:RoundUP(Sum(1 CPU SUs due to first pod * 720hrs * $0.0132 CPU SUs due to extra RAM (8GiB vs 0.4GiB(0.1*4GiB)) * 720hrs * $0.0132 CPU SUs due to more CPU (2vCPU vs 1vCPU(4GiB/4)) * 720hrs * $0.013))=RoundUP(Sum(720(1+2+2)))*0.013$46.80

If you prefer a function for the OpenShift pods here it is:

Project SU HR count = RoundUP(SUM(Pod1 SU hour count + Pod2 SU hr count + …))

OpenShift Pods are summed up to the project level so that fractions of CPU/RAM that some pods use will not get overcharged. There will be a split between CPU and GPU pods, as GPU pods cannot currently share resources with CPU pods.

Storage

Storage is charged separately at a rate of $0.009 TiB/hr

In OpenShift pods, storage is only provisioned while it is active, and in persistent volumes, storage remains provisioned until it is deleted.

Very Important: Requested/Approved Allocated Storage Quota and Cost

The Storage cost is determined by your requested and approved allocation values. Once approved,  OpenShift Resource Allocations, storage quotas are specified by the “OpenShift Request on Storage Quota (GiB)” and “OpenShift Limit on Ephemeral Storage Quota (GiB)” allocation attributes.

Even if you have deleted all volumes, snapshots, and object storage buckets and objects OpenShift projects. It is very essential to adjust the approved values for your OpenShift resource allocations to zero (0) otherwise you will still be incurring a charge for the approved storage. 

Storage Example 1:

• Volume or VM with:500GiB for 699.2hrs
• Will be charged:.5 Storage TiB SU (.5 TiB x 700hrs) x $0.009 TiB/hr$3.15

Storage Example 2:

• Volume or VM with:10TiB for 720hrs (24hr x 30days)
• Will be charged:10 Storage TiB SU (10TiB x 720 hrs) x $0.009 TiB/hr$64.80

Storage includes all types of storage Object, Block, Ephemeral & Image.

SU Conservation – How to Save Cost?

With SUs being the primary metric for resource consumption, it’s crucial to actively manage your workloads when they’re not in use.

Below are practical ways to conserve SUs across different MOC services:

MOC OpenShift

Scale your pods to 0 replicas:

In MOC OpenShift, if your application or job is idle, you can scale its pod replica count to 0. This effectively frees up compute resources (i.e., GPU, vCPU, and RAM) while retaining the configuration, metadata, environment settings, and persistent volume claims (PVCs) for future use.

Using Web Console

1. Go to the MOC‘s OpenShift Web Console.
2. In the Navigation Menu, navigate to the Workloads -> Topology menu.
3. Click the pod or application you want to scale to see the Overview panel to the right.
4. In the Details tab (usually the default tab when you open the deployment):
5. Look for the Pod count or Replicas section.
6. Use the up/down arrows next to the number to adjust the replica count.
7. Set it to 0 by clicking down arrow as shown here:
8. OpenShift will automatically scale down the pods to 0.

When you need to run your application again, you can scale up the pod count or replicas to reclaim the necessary resources.

MOC RHOAI

Toggle the Workbench to “Stopped”:

In MOC Red Hat OpenShift AI (RHOAI), workbench environments can be toggled between Running and Stopped states.

1. Go to the  MOC OpenShift Web Console.
2. After logging in to the NERC OpenShift console, access the MOC’s Red Hat OpenShift AI dashboard by clicking the application launcher icon (the black-and-white icon that looks like a grid), located on the header.
3. When you’ve completed a workload such as model development or experimentation using the Data Science Project (DSP) Workbench, you can stop the compute resources used by the workbench by clicking Stop, next to the Status column for the workbench. Then the status of the workbench change from Running to Stopped, as shown below:This action immediately releases the compute resources allocated to the notebook environment within the Workbench setup.To restart your workbench, click Start next to the Status column for the workbench. The status will change from Stopped to Starting while the server initializes, and then to Running once the workbench has successfully started.
4.