Platform Expertise
Engineers Manage Provisioning, Upgrades, Patching And Monitoring.
We run the control plane, upgrades, patches and monitoring. You deploy. Free HA control plane, bandwidth included, nodes from ₹999 a month, and GPU node groups when your models need them.
Every cluster ships with the parts teams usually spend months building — high availability, upgrades, monitoring, backups and security — already running.
Redundant API servers and etcd across zones on every cluster, at no charge. No per-cluster management fee.
Control plane and worker nodes move to new Kubernetes releases and security patches on a schedule you approve.
Node groups add nodes when pods are waiting for capacity and remove them when demand drops — up to 1,000 nodes.
Node, pod and resource usage in one view, with custom alerts to email, WhatsApp or your on-call tool.
Scheduled backups of cluster configuration and workloads. Restore a namespace or the whole cluster in one step.
RBAC, network policies, encrypted secrets and policy-driven compliance checks, configured from day one.
A clear line between what XcellHost operates and what stays in your hands — so nothing falls between teams.
Traffic enters through managed load balancers and ingress, lands on the right node group, and persists to block storage — with the control plane kept out of your way.
Choose the network provider that fits your policies, and expose services publicly or keep them private.
A default CSI driver gives every cluster persistent block volumes that follow your pods.
Security is configured before your first deployment, not bolted on later.
Run separate node groups inside one cluster, each with its own flavour, size and autoscaling limits.
Scale out fast on load, replace nodes freely. Ideal for horizontal autoscaling.
Steady nodes with persistent block volumes through the CSI driver.
Pack services densely with quotas per namespace and network policies between them.
GPU node groups for model training, inference and data processing, next to your app nodes.
The HA control plane is free and bandwidth isn't metered. Compare that with typical hyperscaler pricing, which bills nodes, the control plane and data transfer separately.
| Worker node | XcellHost / month | Typical hyperscaler* | You pay less | |
|---|---|---|---|---|
| 2 vCPU · 4 GiBDev, test, small apps | ₹999 | −₹1,600 · 62% | ||
| 4 vCPU · 8 GiBMost production apps | ₹1,999 | −₹3,200 · 62% | ||
| 8 vCPU · 16 GiBDatabases, heavy services | ₹3,799 | −₹6,600 · 63% | ||
| GPU node groupTraining, inference, analytics | On quote | — | Sized to your model | |
| HA control planePer cluster | Free | −₹6,499 / cluster |
Indicative · excl. 18% GST. 6-month terms save 5%, 12-month terms save 10% (indicative). *Typical hyperscaler figures are indicative node-only list prices converted to INR; they exclude bandwidth charges.
Pick a node flavour, how many nodes and the term. Add block storage and an autoscaling ceiling to see your peak cost too.
GPU type, count and hours depend on whether you are training, fine-tuning or serving. Tell us the workload and we'll send a sized quote within one business day.
The same five steps the console above walks through. No YAML needed to get started — and full kubectl access once you're in.
Name the cluster and pick a supported Kubernetes version.
Choose Calico, Flannel or Canal, and a public or private API endpoint.
NGINX ingress, default backend, secrets encryption, etcd snapshot backup.
Set flavour, node count, volume size and autoscaling limits.
Confirm and create. Download kubeconfig when the cluster is ready.
Namespaces with resource limits give every squad its own sandbox on shared nodes — without stepping on each other.
Ship new versions to a slice of traffic, watch the metrics, then roll forward or back in seconds.
Load balancing, self-healing pods and horizontal scaling keep busy sites up through traffic spikes.
Deploy, scale and update each service on its own, with policies controlling who talks to whom.
Train and serve models on GPU node groups, scaled to the job and kept next to your data in India.
Burst transcoding and rendering jobs across many nodes, then scale back down when the queue is empty.
It's upstream Kubernetes, so your existing Docker images, Helm charts, Kustomize overlays and CRDs deploy unchanged.
Clusters run in our Indian data centres, including a Tier IV facility, operated by the same certified teams that run XcellHost's managed cloud since 1999. Other regions on request.
Workloads and backups stay in India — helpful for DPDP Act and sector data-localisation needs.
Data encrypted in transit and at rest, with secrets encrypted inside etcd.
Our operations and security centres watch the platform round the clock.
RBAC, policy checks and access logs make audits straightforward. GST invoice on every order.
Engineers Manage Provisioning, Upgrades, Patching And Monitoring.
Run Your Cluster In India, With Other Regions Available On Request.
Plan Containerisation, Workload Migration And CI/CD With Our Team.
A Redundant Control Plane And Self-Healing Workloads.
Operations And Security Teams Support Your Platform Around The Clock.
Cloud Infrastructure, Kubernetes Operations And Support Together.
No. We handle provisioning, upgrades, patching and monitoring of the platform. Your developers deploy apps with the tools they already use, and our engineers are available 24×7 when you need help.
Yes. It's conformant upstream Kubernetes, so Docker images, Helm charts, Kustomize overlays and custom resource definitions (CRDs) deploy without changes.
We run the control plane — API server, etcd, scheduler — plus upgrades, patching and backups. You control node pools, networking choices, autoscaling limits, quotas, namespaces and RBAC.
Every cluster gets a redundant control plane at no charge. Persistent volumes, automated failover, self-healing pods and load balancing keep applications running if a node fails. The platform carries a 99.95% uptime SLA.
You set minimum and maximum nodes per node group. When pods can't be scheduled for lack of capacity, nodes are added; when demand drops, they're removed — up to 1,000 nodes. You pay the node rate for the extra nodes while they run. Use the calculator's peak estimate to budget.
No. The HA control plane is free on every cluster and bandwidth is included rather than metered. You pay for worker nodes and the block storage you provision.
Yes. Our team helps containerise legacy applications, move workloads from other clouds or VMs, and set up CI/CD pipelines with tools such as Jenkins, GitLab CI or Argo CD.
Clusters are hosted in Indian data centres; other regions are available on request. Data is encrypted in transit and at rest, and Kubernetes secrets can be encrypted inside etcd.
Yes. Add GPU node groups to any cluster for training, inference or analytics. GPU capacity is quoted per configuration — request a quote and we'll size it to your model.
In rupees, with a GST invoice. Choose monthly billing or a 6- or 12-month term for an indicative 5% or 10% discount. Pay by UPI, card, net banking or bank transfer.