All systems operational · 99.99% uptime SLA
UK8S — Managed Kubernetes · Dubai · 31 Global DCs · Kubernetes 1.28

Enterprise Kubernetes,
deployed in 5 minutes.

NubexCloud UK8S delivers fully managed Kubernetes — zero-ops control plane, VPC-native pod networking, elastic autoscaling, and GPU support — all backed by 99.95% SLA across UAE and GCC. Focus on your applications, we handle the cluster.

5min
Cluster Provisioning
99.95%
SLA + 100× compensation
K8s 1.28
Latest stable version
$0/mo
Control plane cost
Trusted by teams in
Product Overview

What is UK8S?

UK8S (NubexCloud Container Service for Kubernetes) is a fully managed Kubernetes platform. We provision and maintain the control plane — API Server, Scheduler, Controller Manager, and etcd — at zero cost. You deploy worker nodes and pay only for the compute your workloads actually use.

Free Managed Control Plane

API Server, Scheduler, Controller Manager, and etcd — all managed, monitored, patched, and upgraded automatically by NubexCloud at no charge. Zero ops for your team.

Elastic Node Pools

Add and remove node pools on demand. Mix UHost virtual machines with UPHost bare-metal nodes in one cluster. GPU node pools for AI and ML workloads. Custom images and init scripts supported.

4-Mode Autoscaling

HPA scales pod replicas by CPU/memory or custom metrics. VPA right-sizes resource requests. CronHPA pre-scales on a schedule. Cluster Autoscaler adds/removes nodes based on real demand.

VPC-Native Pod Networking

Pods get real VPC IP addresses — no overlay, no NAT, no encapsulation overhead. Full NetworkPolicy micro-segmentation. Fixed IPs for stateful apps. Seamless ULB load balancer integration.

Multi-Tier Storage (CSI)

Native CSI drivers for UDisk block storage (RSSD NVMe), UFS shared file storage, and US3 object storage. Dynamic PVC provisioning, online volume expansion, and full StorageClass support.

GPU & AI-Ready

Add NVIDIA GPU node pools to any cluster. Run training, inference, and HPC workloads alongside regular pods. GPU sharing across multiple pods. Prometheus-based GPU monitoring built in.

Platform Specs

Built for production-grade scale

Key figures for UK8S clusters running on NubexCloud infrastructure across the UAE and GCC region.

K8s 1.28
Latest Stable Version
UK8S tracks upstream Kubernetes releases. In-place version upgrades from 1.26 → 1.28 without cluster rebuilds. Security patches applied automatically.
99.95%
SLA Uptime
Control plane SLA with 100× compensation. Worker node SLA backed by Tier 3+ / National Standard A data centers. Multi-AZ failover supported.
25 GbE
Node Networking
Worker nodes connect via 25 Gigabit Ethernet with Network Enhancement 2.0 — up to 10 million PPS with dedicated NICs for maximum pod-to-pod throughput.
$0
Control Plane Cost
The managed control plane — API Server, etcd, Scheduler, Controller Manager — is free. Pay only for the worker node compute your applications consume.
Elastic Scaling

Scale without limits

UK8S ships with four complementary autoscaling strategies — use individually or combine them for complete adaptive capacity management.

HPA

Horizontal Pod Autoscaler

Scale pod replicas automatically based on CPU utilization, memory usage, or any custom Prometheus metric. Reacts in seconds to traffic spikes or load drops.

minReplicas: 2
maxReplicas: 20
targetCPU: 75%
VPA

Vertical Pod Autoscaler

Automatically right-size CPU and memory resource requests per pod. Eliminates over-provisioning and OOM kills. Runs in Recommendation, Auto, or Initial mode.

updateMode: "Auto"
minAllowed: 100m/50Mi
maxAllowed: 4/8Gi
CronHPA

Scheduled Autoscaler

Pre-scale pods on a cron schedule before predicted traffic — e-commerce flash sales, business-hour spikes, or batch windows. Works alongside HPA without conflict.

schedule: "0 8 * * 1-5"
replicas: 10
tz: "Asia/Dubai"
CA

Cluster Autoscaler

Add worker nodes when pods can't be scheduled. Remove underutilized nodes after a configurable cool-down. Integrates with node pools for fine-grained control.

minNodes: 2
maxNodes: 50
scaleDownDelay: 10m
Cluster Network

VPC-native pod networking

Unlike overlay-based CNI plugins, UK8S assigns real VPC IP addresses to every pod via IPAMD — delivering the lowest possible latency, no encapsulation overhead, and seamless connectivity with other NubexCloud services.

Underlay network built on UVPC

Every pod receives an IP directly from your VPC subnet — visible to all resources in your network without NAT. The IPAMD daemon manages per-node IP pools while CIDR blocks are pre-reserved to guarantee zero IP exhaustion during rapid scaling.

  • Pods get real VPC IPs — no overlay or tunneling
  • NetworkPolicy for fine-grained pod-to-pod access rules
  • Fixed (static) IP addresses for stateful workloads
  • Hot CNI plugin upgrades with zero service interruption
  • Native ULB integration via Service annotations
  • Nginx Ingress with TLS, path routing, and rate limiting
  • CoreDNS with custom domains and NodeLocal DNS cache
  • kube-proxy in iptables or high-performance IPVS mode
Network Documentation →
Your VPC Network (UVPC) IPAMD Worker Nodes — VPC Pod IPs 10.0.1.4 10.0.1.5 10.0.2.6 GPU Pod NetworkPolicy — micro-segmentation between pods ULB Load Balancer Internal + External Nginx Ingress TLS · Routing · Rate limit Internet →
Cluster Storage

Every storage tier, natively integrated

UK8S ships with CSI drivers for block, file, and object storage — all with dynamic PVC provisioning, online expansion, and full StorageClass support out of the box.

Block Storage
UDisk CSI Driver

Attach NVMe RSSD, SSD, or HDD UDisk volumes as persistent volumes. Dynamic expansion without pod restarts. Automatic pod recovery when a node crashes. Perfect for databases, queues, and high-IOPS workloads.

RSSD NVMe SSD / HDD Dynamic PVC Online Expand
File Storage
UFS (NFS-Compatible)

Shared file volumes mountable by multiple pods simultaneously — ReadWriteMany workloads made simple. Static or dynamic PV provisioning. Ideal for CMS, log aggregation, and ML data pipelines.

ReadWriteMany Dynamic PV NFS-Compatible
Object Storage
US3 (S3-Compatible)

Mount NubexCloud US3 object storage as volumes inside pods. S3-compatible API, unlimited scale, zero pre-provisioning. Ideal for media, backups, static assets, and data lake workloads.

S3-Compatible Unlimited Scale CSI Driver
Observability

Full-stack monitoring & logging

Deploy a production-ready Prometheus stack or centralized ELK log pipeline in minutes — pre-configured for UK8S cluster metrics and logs from day one.

Prometheus + Grafana Monitoring

One-click deploy via UK8S plugin. Scrapes Node, Pod, and container metrics automatically. Pre-built Grafana dashboards for cluster health, resource usage, and network. Alert rules with configurable receivers (email, webhook, Slack). Custom metrics via Prometheus Adapter for HPA scaling targets.

ELK Log Solution

Filebeat DaemonSet collects logs from every node automatically. Ships to Elasticsearch with structured JSON indexing. Kibana dashboards for log search, visualization, and alerting. UK8S Log Plugin enables simplified one-click ELK deployment with lifecycle management built in.

RBAC & API Audit

Fine-grained RBAC Roles and ClusterRoles for every team member. API Server audit logging with configurable verbosity. ServiceAccount token rotation for workload identity. ETCD automatic backup with point-in-time recovery for disaster resilience.

Node Problem Detector

Proactively detects node hardware faults, kernel issues, and runtime errors before they impact workloads. Surfaces problems as Kubernetes Events and Node Conditions. Integrate with your alerting stack for automated incident response and node remediation workflows.

AI & GPU

Run GPU workloads at cloud scale

Add NVIDIA GPU nodes to any UK8S cluster and run AI training, model inference, and scientific computing — with full isolation, sharing, and Prometheus-based monitoring.

GPU+
Dedicated GPU Node Pools
Add GPU-equipped node pools to your cluster with custom node images. NVIDIA device plugin pre-installed — request GPU resources in any pod spec with nvidia.com/gpu resource limits. Mix CPU and GPU pools in one cluster.
÷N
GPU Sharing & Multi-Pod
Run multiple pods on a single GPU card simultaneously using the UK8S GPU-share scheduler. Dramatically reduces cost for inference workloads and small model serving. Each pod receives isolated GPU memory slices and compute fractions with no interference.
📈
GPU Cluster Monitoring
Track GPU utilization, memory usage, temperature, and power consumption per pod in real time via Prometheus metrics and pre-built Grafana dashboards. Set custom alerts to detect GPU saturation, memory pressure, or hardware faults before they impact your AI workloads.
Why UK8S

How UK8S compares

UK8S versus self-managing Kubernetes on cloud VMs — feature by feature.

Feature NubexCloud UK8S Self-managed K8s on VMs Impact
Control plane management Fully managed — free Manual setup & ops Saves 10–20 hrs/month per cluster
K8s version upgrades In-place, 1-click Manual, high-risk Zero downtime upgrades
Pod networking VPC-native, real IPs Overlay (flannel/calico) Lower latency, no encapsulation
etcd backup Automatic, point-in-time Manual scripts required Instant disaster recovery
Cluster autoscaling Built-in (HPA/VPA/CA) Complex custom setup Zero ops elastic capacity
GPU support Native node pools + sharing Manual driver management AI workloads in minutes
Monitoring Prometheus stack, 1-click Manual Helm installs Instant cluster observability
Supported Versions

Kubernetes version support

UK8S tracks upstream Kubernetes releases closely. In-place cluster version upgrades are supported — move from 1.26 to 1.28 without rebuilding your cluster.

Version Status Runtime Notes Release Notes
Kubernetes 1.28 Latest containerd Recommended for all new clusters. Full support. View notes →
Kubernetes 1.26 Maintained containerd Security patches applied. Plan upgrade to 1.28. View notes →
Quick Start

Your first cluster in 4 steps

From zero to a running Kubernetes cluster ready for your first deployment — all within the NubexCloud Console, no CLI required.

1
Create Cluster
Open the UK8S console, choose Kubernetes version (1.28 recommended), VPC, and availability zone. The control plane provisions in under 5 minutes — for free.
2
Add Node Pool
Select UHost instance type, disk type (RSSD recommended), and node count. Configure autoscaling min/max or set a fixed size. GPU node pools available for AI workloads.
3
Connect kubectl
Download your kubeconfig from the console or use Web Kubectl directly in your browser. Run kubectl get nodes to confirm all nodes are Ready.
4
Deploy & Scale
Apply your manifests, Helm charts, or Kustomize configs. Enable HPA, attach persistent volumes, configure Ingress, and go live on NubexCloud infrastructure.
Read the Beginner's Guide →
Documentation

Explore the full UK8S docs

Everything you need to operate UK8S in production — from Day 0 cluster creation to advanced networking, storage, security, and GPU practices.

Get Started Today

Ready to launch your first UK8S cluster?

No credit card required. Free control plane. Pay only for the worker nodes your workloads actually use.

Create Free Cluster View Pricing Talk to Sales
Supported runtimes & integrations
containerd Prometheus Helm Nginx Grafana NVIDIA