Skip to content

Overview

Rafay's GPU Platform as a Service (PaaS)'s goal is to make it easy for data scientists and developers to get access to GPU enabled workspaces where they can perform their tasks quickly and efficiently. Data scientists and developers should not be required to learn about low level infrastructure details and tooling.

The ideal experience for them is when they can get access to GPU resources when they need it and in an environment that has all the necessary libraries, frameworks and applications for their tasks.

GPU PaaS Concept

There are two types of customers/users of Rafay's GPU Platform as a Service (PaaS). Select your option below for more details.

Enterprises

Due to extreme scarcity of GPUs, every organization is now being forced to

  • Make long term commitments to cloud providers, and/or
  • Purchase expensive GPUs (e.g. Nvidia H100) ahead of demand, and/or
  • Source GPUs from multiple cloud providers

Once they get access to the GPUs, IT/Operations struggle to answer the question:

How do we allocate the GPUs to multiple users in a timely and efficient manner?

Benefits

With Rafay's GPU PaaS, Enterprise IT/Operations experience the following benefits:

  • Create and manage a "centralized pool of GPU resources" spanning multiple data centers, clusters and cloud providers.
  • Use their infrastructure (public cloud or data center) and ensure that data never leaves their environment into an untrusted service
  • Do not incur unnecessary data egress costs to 3rd party services
  • Curate and manage organizational standards for workspaces, software frameworks, libraries and integrations with Git
  • Centralized access management to corporate data
  • Manage a GPU pool with heterogenous GPUs (different GPU providers, models)
  • Zero provider lock-in allowing them to operate in any provider of infrastructure of their choice
  • Effortlessly and automatically scale up/down to meet dynamic needs of the organization

With Rafay's GPU PaaS, Enterprise IT/Operations can provide their developers and data scientists with the following benefits:

  • Provide a vending machine style self-service experience to request and use GPU enabled workspaces that can span multiple cloud providers and types of infrastructure (i.e. private data centers and public cloud providers).
  • Deliver a curated, turnkey experience with workspaces that can come preinstalled with critical frameworks, libraries and applications saving them significant time/effort.
  • Ensure that the workspaces are pre-configured with secure access to corporate data and Git

Deployment Options

Rafay GPU PaaS is a control plane product that is available in both SaaS and self hosted software form factors. Every enterprise gets access to an isolated tenant i.e. Rafay Org. Read about the supported multi-tenancy options for GPU workspaces.

GPU PaaS Tenant

GPU Cloud Providers

GPU Cloud providers have to make massive CapEx investments to acquire and operate GPUs in their data centers. Even if they are able to acquire GPUs in a timely manner, they have to address the following challenges:

  • Monetize their significant GPU investment quickly
  • Maximize and optimize GPU utilization to ensure they can deliver very high margins
  • Significant time, effort and deep engineering expertise is required for them to develop/maintain a customer facing platform

Benefits

With Rafay's GPU PaaS, GPU Cloud Providers can:

  • Be operational with a revenue generating GPU Cloud service in a matter of days
  • Avoid spending years developing a complex, multi-tenant, customer facing platform
  • Onboard and manage multiple customers (end users and enterprises) via a multi-tenant platform

Deployment Options

Rafay GPU PaaS is a control plane product that is available in both SaaS and self hosted software form factors. Every GPU cloud provider gets access to a dedicated operations portal and can host/manage multiple enterprises each with their own isolated tenant i.e. Rafay Org.

GPU PaaS Cloud Provider