Deployment models: SaaS, PaaS, on-premise

ZEUS is deployed in three models. The choice depends on data sovereignty, the scope of compliance, and operational capacity — not on feature parity. Each model has the same 22 studios and the same connectors.

Comparison

CriterionCloud SaaSManaged PaaSOn-Premise
Hostingshared ProfessNet clouddedicated ProfessNet VMsthe client's DC
Time-to-value< 30 min1-3 business days1-2 business days
Updatescontinuous CI/CDcoordinated windowquarterly signed tarball
DB isolationshared (Postgres RLS)dedicated DB instance100% sovereignty
Source codenonenoneavailable for audit
Who it's forSMEs, cloud-first startupsbanks, healthcarecritical infra, air-gapped

Cloud SaaS — multi-tenant

ProfessNet hosts the central services; each client has a stateless, read-only connector. Tenants share one hardened backend — data is isolated by Postgres Row-Level Security on every table. Per-tenant subdomains (acme.studio.zeus.io).

Managed PaaS — single-tenant

A dedicated VM cluster per client (backend/frontend/db/redis + separate VMs for the CNAPP scanners). Full isolation: a separate database, backup vault, DNS, Resource Group. Ideal for banks, healthcare, and sovereignty requirements.

On-Premise — air-gapped

The full platform on the client's hardware (Linux + Kubernetes). No outgoing telemetry, no inbound and outbound to ProfessNet. Updates via a quarterly, cryptographically verified tarball. The "Glass Box" model — code available for audit, no lock-in.

Seven layers of isolation (multi-tenant)

In the SaaS model, data isolation is "by construction" — seven independent layers (each self-sufficient): the database (tenant_id filter), the cache (a Redis prefix), secrets (per-tenant Key Vault), the API (a claim in the JWT), studio routing, the ARQ task queue, and a per-tenant hash-chain audit log.

Sales tip: most of the competition is "SaaS only". Our differentiator is on-prem with code audit — crucial for regulated industries and the public sector.


This closes the ZEUS track. After these ten lessons you can explain the platform, run a demo, and operate it as an operator — in each of the three models.