Infrastructure Control Plane

The Vision

A trading operation is only as dependable as the foundation it runs on. The RST Platform can be well designed and carefully operated, and still fail because a disk filled up, a host missed a security patch, or nobody remembers how a resource was configured.

The Infrastructure Control Plane takes responsibility for that foundation, wherever it lives:

• On premises, it covers the physical hardware and the virtualization platform on top of it.
• In the cloud, there is no hardware to look after. It covers the orchestration resource pools and the fundamental parts the platform needs, such as networking, storage, identity, and secrets.

In both cases every resource is declared — what it is, what runs on it, how it is configured — and the control plane brings reality in line with that declaration and keeps it there.

Why This Is Needed

Owning the operation means owning the operations work too — fully on your own hardware, and still substantially in the cloud, where the provider runs the hardware but the configuration, the resource pools, and the recovery are yours. Done by hand, that work drifts: one host is patched, another is not; a configuration change made during an incident is never written down; a restore is assumed to work until the day it is needed.

The answer is not to give up ownership, but to treat the infrastructure as something with a life cycle of its own — designed, provisioned, changed, and eventually replaced — with that life cycle made explicit and repeatable.

What It Covers

Provisioning. Bringing a resource to a known, declared state the same way every time — a physical machine from bare metal, a virtual machine, or a cloud resource pool.

Configuration. Keeping each host’s configuration in version control and reconciling it continuously, so that drift is detected and corrected rather than discovered.

Patching. Applying updates on a schedule, in a controlled order, with the workloads above moved out of the way first — operating systems and firmware on premises, node images and managed service versions in the cloud.

Capacity. Knowing what each resource has, what it is using, and when more is needed — another machine on premises, a larger resource pool in the cloud — before the platform runs out of room.

Recovery. Backups that are tested by restoring them, and a documented, rehearsed path from a failed host or resource pool to a running one.

How It Fits

The Infrastructure Control Plane is the bottom layer. It runs the foundation, on premises or in the cloud; the RST Control Plane runs the RST Platform on them; and the RST Platform runs the trading. Each layer depends only on the one below it, and each can be changed without redesigning the others.

It is built from the same kind of mature, open-source tooling as the rest of the platform, and the decisions behind it are written up in the writings as they are made, starting with I Took Self-Hosting Too Far.