Upstream Kubernetes. We run the control plane.
CNCF-conformant Kubernetes with a highly available control plane we patch and back up, autoscaling node pools, integrated storage and ingress — and no charge for the control plane itself.
- Free HA control plane
- Cluster autoscaling
- GPU node pools
- No proprietary APIs
Kubernetes pricing
The control plane costs nothing. Worker nodes are billed at standard instance rates, so a cluster costs the same as the servers underneath it.
| Component | Detail | Price (AED) |
|---|---|---|
| Managed control plane (HA) | 3 replicas, patched and backed up | 0 |
| Worker node — 4 vCPU / 8 GB | Standard instance rate | 189 / month |
| Worker node — 8 vCPU / 16 GB | Standard instance rate | 369 / month |
| Worker node — 8 dedicated / 32 GB | Dedicated core instance rate | 1,049 / month |
| GPU node — RTX PRO 6000 | Per GPU, on demand | 5.50 / hour |
| Persistent volumes (CSI) | Block storage rate | 0.36 / GB / month |
| Ingress load balancer | LB-Standard rate | 79 / month |
Prices exclusive of 5% UAE VAT. Cluster autoscaling adds and removes nodes automatically; you are billed hourly for the nodes that actually run.
The control plane work you stop doing
Nothing here is impossible to run yourself. It's the ongoing maintenance that most teams would rather not own.
| Aspect | Self-managed | CloudTechtiq managed |
|---|---|---|
| Control plane setup | You provision etcd, API servers and networking by hand | Live in minutes, HA by default |
| Version upgrades | Scheduled downtime, manual etcd backup first | Rolling upgrade, zero control-plane downtime |
| Control plane cost | Runs on nodes you pay for and patch | Free — you pay for worker nodes only |
| Storage integration | CSI driver installed and maintained by your team | Preinstalled, snapshot support out of the box |
| Autoscaling | Cluster autoscaler configured and monitored by you | One toggle, tuned to the node pool limits you set |
| 3am etcd corruption | Your incident | Our incident |
What is included in the cluster
The components most teams end up installing anyway, configured and maintained rather than left to you.
HA control plane
Three API server replicas with etcd backed up on a schedule, upgraded in place without cluster downtime.
Cluster autoscaler
Node pools grow when pods are unschedulable and shrink when they are not needed, within limits you set.
CSI storage driver
PersistentVolumeClaims map to block volumes automatically, with snapshot support through the standard API.
Ingress and service load balancing
Service type LoadBalancer provisions a real load balancer; NGINX ingress is available preconfigured.
GPU node pools
NVIDIA device plugin ready, so inference and training pods schedule onto GPU nodes in the same cluster.
RBAC and network policy
Role-based access wired to your identity provider, plus CNI-enforced network policy between namespaces.
From git push to running pods
The cluster ships with the hooks GitOps controllers expect, so deployments stay declarative from day one instead of turning into ad-hoc kubectl sessions.
Push to git
A commit lands on main, or a PR merges — your Git provider is the single source of truth for what should be running.
Image builds in CI
Your existing pipeline builds and pushes a container image, tagged with the commit SHA, to your registry of choice.
Argo CD or Flux syncs state
The GitOps controller running in-cluster notices the manifest change and reconciles the cluster to match it automatically.
Rollout completes
Pods roll out with the readiness checks you defined. If a rollout fails, roll back is a git revert, not a manual kubectl scramble.
Frequently asked questions
Yes. You pay for worker nodes, block volumes and load balancers at their normal rates. The control plane — API server, scheduler, etcd — is run, patched and backed up by us at no charge.
Upstream, CNCF-conformant Kubernetes. Standard manifests, Helm charts and operators work without modification, and there is no proprietary API you would have to unwind if you moved elsewhere.
You choose the target version and a maintenance window. The control plane upgrades first, then node pools roll one node at a time with cordon and drain, so a correctly configured deployment stays available throughout.
Yes. GPU node pools with the NVIDIA device plugin pre-installed attach to the same cluster, so inference services schedule onto GPU nodes while the rest of the platform runs on standard nodes.