Aller au contenu principal

Hybrid Deployment

STOA Platform supports multiple deployment models to match your security, sovereignty, and operational requirements.

Deployment Models

ModelControl PlaneGatewayData ResidencyBest For
HybridSTOA Cloud (EU)On-PremisesBusiness data on-premMost enterprises
Full On-PremisesYour infrastructureYour infrastructureFull on-premMax sovereignty
Multi-CloudSTOA CloudMultiple regionsDistributedGlobal organizations

Control Plane Cloud + Gateway On-Premises

The default deployment model balances ease of management with data sovereignty.

What Stays On-Premises

Data TypeDescriptionEncryption
API PayloadsRequest/response bodiesTLS in transit
CredentialsAPI keys, tokens, certificatesAES-256 at rest (Vault)
User IdentitiesOracle OAM/OIM directoryExisting controls
Raw LogsFull transaction detailsCustomer-controlled
SecretsHashiCorp Vault dataAES-256-GCM

What Goes to Cloud

Data TypeDescriptionSensitivity
API MetadataNames, descriptions, OpenAPI specsLow
Aggregated MetricsRequest counts, latencies, errorsLow
ConfigurationRouting rules, policiesLow
Federated TokensShort-lived, no credentialsLow

Network Requirements

DirectionProtocolPortsPurpose
On-prem → CloudHTTPS443Config sync, metrics push
Cloud → On-premNoneNo inbound required

Key security benefit: No inbound connections required. All communication is initiated from your infrastructure.

Prerequisites

  • Kubernetes 1.28+ cluster on-premises
  • Outbound HTTPS to STOA Cloud endpoints
  • DNS resolution for STOA services
  • Existing identity provider (OAM, Okta, Azure AD)

Model 2: Full On-Premises

Maximum Sovereignty

For organizations requiring complete control over all components.

When to Choose Full On-Premises

  • Regulatory requirement for 100% data residency
  • Air-gapped environments
  • Government or defense sector
  • Extreme compliance requirements (banking regulators)

Additional Requirements

ComponentOn-Premises Requirement
KubernetesProduction cluster (3+ nodes)
PostgreSQLHA setup (primary + replica)
OpenSearchCluster for logs and search
Object StorageS3-compatible (MinIO)
TLS CertificatesPKI or Let's Encrypt

Trade-offs

AspectHybridFull On-Prem
Operational complexityLowerHigher
UpdatesAutomaticManual
SupportFullLimited self-service
Data residencyMetadata in cloud100% on-prem
Initial setup1-2 days1-2 weeks

Model 3: Multi-Cloud (Future)

Global Distribution

For organizations requiring presence in multiple regions or clouds.

Planned Capabilities

  • Geographic load balancing
  • Data residency per region
  • Cross-region failover
  • Unified global dashboard

Status: Roadmap Q4 2026


Data Flow Summary

Hybrid Model Data Matrix

DataDirectionEncryptionFrequency
ConfigCloud → On-premmTLSOn change
Metrics (aggregated)On-prem → CloudTLSEvery 15s
AlertsCloud → Ops teamTLSOn trigger
Tokens (federated)Cloud ↔ On-premTLSPer request
PayloadsDesigned to remain on-premN/AN/A
CredentialsDesigned to remain on-premN/AN/A

Network Diagram


Getting Started

Hybrid Deployment Quick Start

Private Beta

Repository access is granted to beta participants. Request access to get the Helm chart and deployment instructions.

# 1. Create namespace
kubectl create namespace stoa-system

# 2. Add the STOA Helm repository (provided with beta access)
helm repo add stoa https://charts.gostoa.dev
helm repo update

# 3. Install with hybrid configuration
helm install stoa stoa/stoa-platform \
--namespace stoa-system \
--set mode=hybrid \
--set controlPlane.endpoint=https://api.<YOUR_DOMAIN> \
--set controlPlane.tenantId=YOUR_TENANT_ID

# 4. Verify installation
kubectl get pods -n stoa-system

Full On-Premises Quick Start

# 1. Install prerequisites
helm install postgresql bitnami/postgresql -n stoa-system
helm install opensearch opensearch/opensearch -n stoa-system
helm install vault hashicorp/vault -n stoa-system

# 2. Install STOA full stack
helm install stoa stoa/stoa-full \
--namespace stoa-system \
--set mode=on-premises \
--values your-values.yaml

Next Steps


Need help choosing the right deployment model? Contact us for an architecture consultation.