Private Cloud

General definition

A Private Cloud is one of the deployment models defined in NIST Special Publication 800-145: cloud infrastructure operated solely for one organisation, whether it owns it, leases it or has a third party manage it, and whether it sits on-premise or in a provider’s facility. What makes it a cloud rather than a traditional server room is the operating model: virtualised or containerised resources, self-service provisioning, automation and measured use. What makes it private is that no other customer shares the underlying hardware and network.

• Public cloud: shared infrastructure sold on demand by AWS, Google Cloud, Azure and others; the customer isolates itself with a virtual private cloud
• Private cloud: dedicated infrastructure, built on platforms such as VMware, OpenStack, Nutanix or a Kubernetes distribution, or delivered as hosted private cloud by a provider
• Hybrid cloud: private and public environments joined so workloads and data can move between them under policy
• Dedicated instances in public cloud: single-tenant hardware inside a public provider, a middle path that some regulators accept as equivalent

Organisations choose private cloud for control rather than price. Regulated sectors need to state where data is stored and who can reach the hardware; some jurisdictions and contracts require data to stay in-country or in-house; latency-sensitive or data-heavy workloads can be cheaper on owned capacity; and security teams prefer a perimeter they fully instrument. The cost is operational: someone has to run the platform, patch it, back it up and plan capacity, which is why hosted and managed private clouds exist.

For application vendors the private cloud question is really a deployment question. Software that only runs as the vendor’s multi-tenant SaaS cannot be placed in a private cloud at all. Software that ships as containers with documented operations, or that the vendor will install and run inside the customer’s environment, can. That distinction is what separates an on-premise or single-tenant capable product from one that merely claims enterprise readiness.

In the Ethora ecosystem

Ethora is built to be placed. The open-source server ships as a Docker deployment with Grafana and Prometheus monitoring and load-testing tools, so a customer’s platform team can run it in a private cloud built on VMware, OpenStack or Kubernetes exactly as they run their other workloads. The dedicated server option adds operations: Ethora TechOps install and run the platform inside the customer’s own cloud account or on-premise environment, single-tenant, in the region the customer specifies.

Either way the messaging data, files, call media and AI agent knowledge bases stay on infrastructure dedicated to the customer, which is the basis for the BAA and DPA Ethora signs on those plans and for EU-only residency where it is required. AI agents can use self-hosted models on the same private infrastructure, so a healthcare or finance deployment need not send transcripts to any third-party model API.