How Deloitte Cut Environment Provisioning Time by 45+ Minutes with vCluster

Provisioning Was Slow, Costly, and Dependent on Platform Engineers

Before vCluster, Deloitte provisioned dedicated EKS clusters on AWS for each ephemeral testing need—an approach that could take up to an hour per cluster. QA engineers required isolated environments to test specific combinations of application components, but they relied heavily on the platform team to manually spin up and manage those clusters. Each environment also carried the overhead of its own ingress controllers, DNS setup, and monitoring agents—creating significant infrastructure duplication and operational load.

• 45–60 minute wait times to provision each EKS cluster
• Heavy platform team involvement for every environment request
• Duplicated platform stacks (NGINX, DNS, metrics) across clusters
• Wasted compute and memory across isolated EKS clusters
• High operational overhead for upgrades and troubleshooting

Fast, Self-Service Ephemeral Environments with Zero Kubernetes Overhead

After adopting vCluster Enterprise, Deloitte’s platform team built a seamless, self-service interface on top of vCluster’s API. QA engineers can now launch disposable Kubernetes environments in under 3 minutes—without touching a single kubectl command. “Today, QA engineers don’t need to know Kubernetes at all. They just click a few buttons, and a complete environment is up and running in minutes,” said Samuel Lefki, Platform Infrastructure Lead at Deloitte. 

These virtual clusters are automatically registered in Rancher, integrated into Deloitte’s CI/CD pipelines, and provisioned on a shared 12-node EKS host cluster running 25–30 virtual clusters at any given time. Deloitte uses vCluster’s templating system to ensure every environment is provisioned consistently, reducing configuration drift across teams. SSO integration secures access to these clusters without introducing new identity systems, while maintaining tight role-based controls.

vCluster’s reverse proxy, external DNS syncing, and templating features reduced the need for duplicated tooling, slashed compute usage, and simplified support. Engineers now resolve most issues by simply tearing down and spinning up a new environment—without waiting for platform support.

• Run 30+ virtual clusters efficiently on a single shared EKS host cluster
• Save 30+ CPUs, 120+ GB of memory and 60% reduction in node usage by consolidating infrastructure
• Enable 100% self-service access to environments for QA and app teams
• Resolve issues up to 2 hours faster with disposable clusters
• Replace 10 tool deployments with a single shared stack

“Developers care about getting their applications up and running fast. With vCluster, they click a few buttons and get a fully working environment with a reachable URL in minutes. That kind of speed was unheard of before vCluster,” said Ismail A Mohammedali, Platform Engineer at Deloitte.

Balancing Speed, Isolation, and Simplicity

Deloitte evaluated other options like namespace isolation and alternate automation tooling but chose vCluster for its unique ability to deliver true isolation, fast provisioning, and full Kubernetes compatibility. vCluster offered:

• Easier Integration: Seamless compatibility with Rancher and GitHub-based CI/CD pipelines
• True Environment Isolation: Instant access to fully isolated, ephemeral Kubernetes clusters
• Lean, Centralized Stack: Shared platform components like NGINX, DNS, and monitoring agents run once on the host cluster

Deloitte is now exploring broader adoption of vCluster beyond QA, with other teams starting to spin up environments independently. As they continue to streamline operations and integrate more tooling, vCluster is poised to support additional use cases and scale with their evolving platform needs. Features like SSO and templates have made it easier for Deloitte to extend these environments securely, efficiently, and consistently across the organization.