Serverless Pods
Rafay’s Serverless Pods simplify access to compute so your developers and data scientists can focus on innovation, not infrastructure. Instead of wrestling with Kubernetes manifests, namespaces, or ingress configurations, teams can instantly deploy catalog-driven or custom workloads through a governed, self-service experience. The result is faster time-to-value, higher productivity, and a cloud-like developer experience.
With Rafay’s Serverless Pods, data scientists can run notebooks and inference workloads instantly, developers can provision familiar OS environments with preloaded libraries, all through a reusable set of templates that streamline operations and reduces complexity.
.png)
How do serverless pods work?
Serverless Pods are Kubernetes-native environments that act like lightweight, on-demand virtual machines. Each pod runs inside a dedicated namespace that is secured from lateral escalation with the necessary lifecycle management, credential handling, and access controls automatically configured. Instead of creating separate templates for Notebooks, inference services, or developer VMs, Rafay customers can now use one reusable template—the Serverless Pods template, to support all these use cases.
Why Choose Rafay Serverless Pods?
Although there are many situations where offering full or partial access to the Kubernetes cluster makes sense, developers usually just want to deploy their workload and get going, and not be forced to deal with Kubernetes deployments, per-service templates, service accounts, etc. With Rafay’s Serverless Pods offering, developers experience:
Speed
Instant service delivery, faster onboarding
Overall Simplicity
Less code, fewer templates to manage
User Consistency
One experience for all users
Seamless Integration
Deeply tied into Rafay’s inventory and governance controls
Enhanced Security
namespaces, RBAC, and audit logs ensure enterprise readiness
How do serverless pods work?
Serverless Pods are designed specifically for Kubernetes (K8s) form factor deployments and operate on the concept of a host cluster.
Leveraging our deep Kubernetes expertise, we envisioned a solution where K8s pods could provide the desired operating system and pre-installed software, coupled with an integrated SSH experience. To achieve this, we developed a serverless-pods workflow handler that orchestrates the creation of all necessary K8s resources.
Request → Provisioning
- When a user requests a pod, Rafay’s workflow handler automatically provisions all required Kubernetes resources.
- A namespace is created for tenant isolation, a deployment manages the workload, and a secret securely stores SSH keys.
Access Options
- For SSH access, Rafay provisions a NodePort service, enabling direct terminal connections.
- For web-based access (e.g., Jupyter Notebook), Rafay exposes a secure URL via Ingress.
SKU-Based Customization
- Administrators configure SKUs in the PaaS studio, specifying the host cluster and access policies.
- Any Docker image can be used, making the template image-agnostic.
Serverless Pods as a Standardized Template
Every modern application needs a landing zone to run, so why is the process of provisioning landing zones so complicated? Enterprises that streamline the process of setting up landing zones by providing self-service for developer and cloud operation teams gain significant benefits.
Use one versatile template
for developer pods, Jupyter Notebooks, inference services, and beyond.
Accelerate delivery
new SKUs can be created in hours, not weeks.
Reduce maintenance
one well-understood pattern eliminates template sprawl.
Enable flexibility
teams bring their own Docker images and simply attach them at the SKU level.
Match workloads to inventory automatically
with the Rafay Platform's built-in inventory and matchmaking system.
From AI PODs to GPU Cloud: How Cisco and Rafay Deliver Production-Ready, Multi-Tenant AI Infrastructure
Learn how Cisco UCS X-Series + Intersight enable high-performance GPU infrastructure and how Rafay transforms that into a self-service, multi-tenant GPU Cloud in just weeks.
