Skip to content

Oct

Info

GPU PaaS releases are initially rolled out via Rafay's Air Gapped Controller form factor. These will be periodically bundled and rolled out into Rafay's Production SaaS.

v3.1-36

08 Oct, 2025

Integrated GPU Dashboard for Kubernetes Clusters

Users that launch/use Kubernetes cluster based compute instances are now provided seamless, 1-click access to an integrated GPU Dashboard for all GPUs attached to the Kubernetes cluster.

GPU Dashboard

Once users click on the GPU Dashboard, they are presented with the list of GPUs

GPU List

When they select a specific GPU by selecting "View GPU", they are presented with a detailed GPU Dashboard with time series data presenting critical metrics.

View GPU Metrics


Display Type Selector for SKUs

SKU Designers in the PaaS Studio now have the ability provide users with an Increment/Decrement type field for user input. For example, when accepting user input for storage, the admin may wish to force the user to select storage in increments of 5GB as shown in the image below.

Increment/Decrement Example

In PaaS Studio, for a specific input, admins can select "increment/decrement" from the type, provide the step value (e.g. 5GB), a minimum and max value and specify the units.

Increment/Decrement Type


Estimated Costs in Hourly vs Monthly Units

SKU Designers using the PaaS Studio can now specify the display units for the estimated costs shown to users before a compute/service instance is provisioned. They can specify the units to be displayed either as "monthly" or "hourly".

Display Units - Monthly

- name: medium-vllm-inference
  quota:
    max-instances: 5
  billing:
    estimationUnit: monthly
    currency: USD
  dimensions:
    - instance
  ratecard:
    instance:
      - price: 6
        time_unit: h
        currency: USD
        base_unit: 1
        alias: INSTANCE

In the example YAML shown above, the estimationUnit is specified as "monthly". If this were switched to "hourly", the user will be shown the hourly units as shown in the example below.

Display Units - Hourly


Usage Metrics - Filter by "Running" State

Service Providers can now filter compute and service instances by their "running" state in the usage metrics dashboard.

Filter by State for Usage Metrics


Custom Apps in App Switcher

The default portal switcher in the Rafay Org provides links to the Infra, PaaS Studio and end user facing self service portal (developer hub).

Default App Switcher

The portal switcher can now be personalized by users with the superadmin role. They can add "custom applications" that are managed outside the Rafay Platform providing users with a seamless access experience to the entire portfolio of applications supported by the provider. The apps can be presented either in grid or list format to the users.

Custom Apps


Environment Variables in Global Settings

Instead of managing environment variables in the Global Settings YAML file, administrators can now manage these as separate environment variables. These will be automatically appended to Global Settings as overrides.

Env Vars in Global Settings

Admins can now manage sensitive data such as API Keys etc as environment variables

Add Env Variables


Default Billing Currency

Admins using the Ops Console can now specify the default billing currency for customer orgs (tenants).

Default Billing Currency


Disable Launch of New Instances

Admins can disable (via the Ops Console or programmatically via APIs) users from launching new instances in a Customer Org . Users can be shown a customizable message when they attempt to launch new instances. Typical use cases for this are:

  1. KYC of Customer Org needs to be performed before they are allowed to launch/use instances
  2. Customer delinquent with payments

Disable New Instance Launch


Common IDP for All Customer Orgs

Service Providers can configure a centralized Identity Provider (IdP) in the Default Org where they can manage identities for all their customer orgs. Authentication of users across all customer orgs will be performed by the centralized IdP.

Common IDP


Datacenter Configuration for Compute Profiles

Admins managing a compute profile (SKUs) can specify "datacenter" and "storage" infrastructure configuration. For example, they may wish to pin a compute profile to be deployed only in a specific datacenter. They can also optionally make this visible to end users and also expose this as an override for users to select from a dropdown.

Infra Configuration


Auto Share SKUs with All Projects

Profiles (SKUs) shared with a customer org will be automatically made available to users in ALL Projects in the Org and do not require intervention by the Org Admin. This behavior can be optionally disabled by the service provider upon request by the Org Admin.

Infra Configuration