Self-managed Kubernetes on Worldstream: choosing your infrastructure
Kort antwoord
Running your own Kubernetes cluster, with kubeadm, k3s or a similar tool, is a well-established pattern on rented infrastructure, and it works on Worldstream today using products that already exist. Bare Metal Compute gives you the most raw performance and no noisy-neighbour risk. Flexible VPS is the easiest to add nodes to and resize. Flexible Cloud fits best if you're already running VMware and want your cluster to sit alongside it. None of this depends on Worldstream's managed Kubernetes product, which is still preview and not covered here.
Why self-managed, and why now
Plenty of teams run Kubernetes on infrastructure that was never sold to them as "Kubernetes infrastructure": a handful of VPS instances, a rack of bare-metal nodes, or VMs inside an existing private cloud. You install kubeadm, k3s, or another distribution yourself, and you own the cluster lifecycle, upgrades, and node management end to end. It's a common pattern with providers that don't (yet) offer a managed control plane, and it remains a reasonable choice even once one exists, since it gives you full control over versions, add-ons, and node placement.
Worldstream doesn't have a published, self-service managed Kubernetes product today. What it does have is three compute products you can build a self-managed cluster on top of, each with a different trade-off. This article compares them so you can pick a starting point.
The three infrastructure choices
Bare Metal Compute. Dedicated physical nodes with no hypervisor layer between your cluster and the hardware. This gets you the best raw performance and rules out noisy-neighbour effects entirely, since nothing else runs on the box. It's the right starting point if your workloads are CPU, memory, or I/O intensive, or if you're already choosing bare metal for other reasons. See Bare Metal Compute vs. Dedicated Servers for how that product fits alongside Worldstream's other physical-server line.
Flexible VPS. Virtual servers you create from Portal and resize as the cluster's needs change. It's the easiest option to grow: adding a node, or resizing an existing one, involves no hardware lead time. The trade-off is that you're sharing the underlying physical infrastructure with other tenants, so isolation is weaker than bare metal. See Deploying your first Flexible VPS for the provisioning basics.
Flexible Cloud. Flexible Cloud is Worldstream's VMware-based platform, built on VMware Cloud Foundation (vSphere, NSX-T and all-flash NVMe vSAN) and managed through VMware Cloud Director. If your organisation already runs infrastructure there, running Kubernetes nodes as VMs alongside it keeps the cluster inside the same VMware tooling, networking and operational processes you already use for everything else. It's less about raw performance and more about not introducing a second infrastructure silo. See Managing your Flexible Cloud environment via VMware Cloud Director for what that environment looks like day to day.
Comparison
| Bare Metal Compute | Flexible VPS | Flexible Cloud | |
|---|---|---|---|
| Isolation / noisy-neighbour risk | None, dedicated physical hardware | Shared underlying physical infrastructure | Your own VMware environment on Worldstream's VMware Cloud Foundation platform |
| Resize flexibility | Lowest, adding capacity means adding or replacing physical nodes | Highest, resize or add nodes from Portal | Add or resize VMs within the resources allocated to your environment |
| Best fit | Performance-sensitive or I/O-heavy workloads, licensing that rules out virtualisation | Test clusters, early production, workloads that change size | Teams already standardised on VMware tooling and workflows |
One more option sits outside this comparison. If the deciding factor is that the infrastructure must not be shared with other customers at all, Private Cloud is Worldstream's dedicated, non-shared environment on its own network. See Private Cloud vs. Flexible VPS for how that compares with the shared platform.
Picking a starting point
If you're not sure yet, Flexible VPS is the lowest-friction way to stand up a first cluster and learn what your workload actually needs before committing to bare metal or a VMware environment. You can always migrate nodes to a different infrastructure product later, since the cluster itself, and everything you deploy into it, doesn't care what the underlying nodes are running on. See Do you need Kubernetes? A decision guide if you haven't settled the bigger question yet, and Sizing a Kubernetes cluster for high availability once you're ready to plan node counts.