Neoclouds and GPU clouds
Kubernetes for every customer, without building it.
For neoclouds and GPU cloud providers. Give each customer a dedicated, highly available Kubernetes control plane while their workloads run on your hardware.
The problem
1.
Customers expect Kubernetes
Teams renting GPUs expect a Kubernetes API to schedule onto, not only servers to SSH into.
2.
Shared control planes leak
One control plane for many customers means shared etcd, shared API limits, and one tenant’s load in everyone’s cluster.
3.
Building it takes a team
A control plane per customer means running many etcd clusters, certificate authorities, and upgrades at once.
What Kestrel runs for you
A dedicated, highly available Kubernetes control plane. Your workers stay on your nodes.
One control plane per customer
Each customer gets an API server, etcd cluster, and certificate authority of their own. Nothing is shared between tenants.
Your hardware, your customers
Workers stay on your GPU servers. We never host your customers’ workloads.
Highly available
Control planes run across failure zones, so one zone going down doesn’t take your customers’ clusters with it.
Help with the rollout
Architecture and security reviews as you bring Kubernetes to your customers.
Questions
Things people ask us.
Can each of our customers get their own control plane?
Yes. Every control plane has its own API server, etcd cluster, and certificate authority, so customers are isolated from each other.
Do you host our customers’ workloads?
No. Workloads run on your servers. We run only the Kubernetes control planes.
How do we get started?
Join the waitlist or email partnerships@kestrelinfrastructure.com. We’re onboarding a few teams at a time.
Also built for
- AI and GPU clustersGPU clusters for training and inference, with the control plane off your GPU servers.
- Bare metal KubernetesYour own servers, in your data centers or colocation, without operating etcd.
- Data centers and colocationKubernetes on the servers in your facility, without hiring a platform team.