Linode Kubernetes Engine

What Linode Kubernetes Engine actually is

Linode Kubernetes Engine (LKE) is the managed Kubernetes service that grew out of Linode, the Philadelphia-based cloud founded in 2003 and folded into Akamai after the 2022 acquisition. LKE launched on 19 November 2019, and the pitch has not changed since: you get a managed, CNCF-certified control plane and pay nothing for it — the bill is your worker nodes, storage and load balancers. In our view that pricing model is still the cleanest in the managed Kubernetes market, and it is the reason LKE stays on our shortlist in our Kubernetes comparisons.

The product ships in two tiers. Standard LKE gives you a free, single-tenant-of-one control plane, node pools, autoscaling, automatic Kubernetes version updates, and up to 250 worker nodes with the Calico CNI. LKE Enterprise is the newer tier: a dedicated, always-HA control plane at a flat $300/month per cluster, up to 500 nodes, Cilium instead of Calico, VPC isolation and premium load balancers. Enterprise is sold with limited availability in select regions, so treat it as an upgrade path rather than a starting point.

Operationally, LKE behaves like the other developer clouds. You create a cluster from Cloud Manager, the CLI, the API or Terraform, pick node pools from the ordinary compute plan catalog, and the control plane handles etcd, upgrades and node recycling for you. Version management is hands-off in the right way: minor Kubernetes releases roll out through automated, zero-downtime updates, and node recycling means you are not babysitting kubelet drift across the fleet. Because LKE is CNCF-certified and upstream-compatible, manifests, Helm charts and operators that work on GKE or EKS redeploy here without edits — Akamai publishes a GKE migration guide, and we view the portability story as one of the stronger arguments for the platform. Bring your own tooling: kubectl, Rancher, Argo, Terraform and the Akamai CLI all plug in through open APIs. Akamai also sells an "App Platform for LKE" add-on that bolts on Argo CD, Harbor, Prometheus, Grafana, Loki, cert-manager and Istio as a pre-tested stack. That matters if you want a platform team experience without assembling the pieces yourself.

Pricing: what you really pay

The control plane is free. Full stop — no per-cluster fee, no per-hour control plane charge. Your cluster cost is the sum of its worker nodes, and those are billed at standard Akamai Cloud plan prices. The floor is the Nanode 1 GB plan at $5/month per node (1 vCPU, 1 GB RAM, 25 GB storage, 1 TB transfer), so a one-node throwaway cluster costs $5/month and the smallest realistic three-node shared cluster is listed at $36/month total (3 vCPU, 6 GB RAM, 150 GB storage, 6 TB transfer across the pool). Dedicated-CPU three-node clusters start at $108/month on G6 or $129/month on G7 hardware; a three-node high-memory cluster starts at $180/month.

The control plane is only free in its basic form. High availability costs $60/month per cluster and is billed per region. LKE Enterprise costs $300/month per cluster on top of every node. Beyond compute, expect NodeBalancer charges, block storage, and egress at $0.005/GB once you pass the bundled transfer allowance — and note that the newest G8 plan generation bundles no transfer at all, so egress is metered from the first byte there.

Billing is hourly up to a monthly cap: a 1 GB VM accrues $0.0075 per hour and stops at the monthly figure. Backups are not included — Akamai Backups is a paid add-on starting at $2/month for a 1 GB instance. There is no free tier; new accounts occasionally get promotional credit, but nothing permanent.

One trap deserves a paragraph: enabling HA on an existing cluster is irreversible. Akamai's docs state the upgrade deletes all nodes and recreates them, erasing local storage such as hostPath volumes, and the cluster cannot be downgraded afterwards. Decide on HA at creation time rather than learning this during a migration.

Performance & infrastructure

Worker nodes are ordinary Linode plans, so the performance profile is whatever you pick: shared vCPU for dev and light services, dedicated vCPU (G6/G7/G8 on AMD EPYC) for steady loads, high memory for caches and databases, and GPU pools for inference and training. The GPU lineup is current-generation — NVIDIA RTX 4000 Ada nodes from $0.52/hour and RTX PRO 6000 Blackwell nodes up to eight GPUs per node — which puts LKE in the running for AI workloads, not just web services. Node pools support autoscaling, and pods that fail are rescheduled automatically.

On footprint, third-party counts put Akamai Cloud at 33 core compute regions across six continents, including roughly ten in Europe (Amsterdam, Frankfurt, London, Madrid, Milan, Paris, Stockholm and others), Toronto and São Paulo, plus Chennai, Mumbai, Singapore, Tokyo, Osaka, Jakarta, Sydney and Melbourne in Asia-Pacific. LKE runs in core compute regions only — the newer distributed regions support a reduced service set and are not on the LKE menu yet. Every compute instance gets IPv6 by default (a SLAAC address, with routed /64 and /56 ranges available), and Clusters inherit the underlying network, including private networking and Cloud Firewall.

Reliability deserves a sober read. The standard control plane is not replicated: Akamai's own documentation is explicit that by default the components are singletons and a failure "may cause issues with your cluster." The HA option triples etcd and the API server and duplicates the rest with leader election, which is where the published 99.9% uptime guarantee for the control plane and worker node API comes from. LKE Enterprise carries the same 99.9% SLA with dedicated resources. On standard LKE, the guarantee is best-effort. That is the trade you make for a $0 control plane.

Where it falls short

The feature set is thinner than EKS, GKE or AKS. There is no equivalent of GKE Autopilot, no first-party managed ingress/observability bundle included by default (the App Platform add-on is the answer, and it costs extra), and enterprise controls like VPC isolation and Cilium are locked to LKE Enterprise. Enterprise availability is limited and onboarding goes through sales.

The HA story is awkward for existing clusters: the $60/month option sounds cheap until you price in the one-way upgrade that rebuilds your nodes. Teams that started small on standard LKE and later need HA face real migration work.

Egress economics have drifted. Bundled transfer on older plans was generous, but G8 plans bundle none and bill at $0.005/GB, and region pricing varies — Jakarta and São Paulo run higher. For traffic-heavy workloads those two facts can dominate the invoice more than node price does. Check the numbers against our benchmark data before you commit.

Support is tiered now that Akamai owns the platform. Standard plans get ordinary Akamai support; the hands-on managed service is a paid product. If you were used to Linode's old "we'll help anyone" posture, calibrate expectations.

Who should pick this

We'd pick LKE when you want production Kubernetes and the control plane cost of EKS/GKE (which charges per cluster-hour or per cluster-month) is genuinely annoying — small clusters, bursty dev/test environments, and multi-cluster fleets where per-cluster fees multiply. It is also a strong home for GPU inference clusters, because the node pricing is hourly and the GPU catalog is current.

We'd skip it for workloads that need a guaranteed control plane from day one (spec HA at creation), for shops that want Autopilot-style nodeless abstractions, or for organizations that require the compliance surface and regional depth of the hyperscalers. It also is not a substitute for a managed database or object storage layer — those are separate Akamai products with separate bills, so factor them into the architecture rather than assuming they ride along with the cluster. As always, cross-check node pricing and transfer allowances against our benchmark numbers before you size a production cluster; the plans move faster than most review roundups do.

Key takeaways

• Free managed control plane on the standard tier; you pay only for worker nodes, storage and load balancers.
• HA control plane is $60/month per cluster and enabling it on an existing cluster is irreversible (nodes rebuilt, local storage erased).
• LKE Enterprise adds a dedicated HA control plane, Cilium and VPC isolation for $300/month per cluster, with limited availability.
• Entry cost is $5/month per Nanode 1 GB node; egress is $0.005/GB and the newest G8 plans bundle no transfer.
• Standard-tier control plane is not replicated and carries no SLA — 99.9% comes with HA or Enterprise.
• Broad footprint (33 core regions on third-party counts) with IPv6 on every instance and hourly billing up to a monthly cap.

Sources

• https://www.akamai.com/products/kubernetes
• https://techdocs.akamai.com/cloud-computing/docs/high-availability-ha-control-plane-on-lke
• https://techdocs.akamai.com/cloud-computing/docs/linode-kubernetes-engine
• https://www.linode.com/pricing
• https://www.linode.com/products/backups
• https://hostingengines.com/linode-pricing/
• https://comparedge.com/tools/linode/deployment-regions
• https://www.prweb.com/releases/linode-launches-managed-kubernetes-engine-848995007.html
• https://linode.com/docs/products/compute/compute-instances/guides/ipv6