Deployment

Turqoa supports three deployment models to accommodate different infrastructure requirements, security postures, and operational constraints. This guide helps you choose the right model and understand the architecture of each.

Deployment Models

Model	Description	Best For
On-Premise	Full Turqoa stack runs on hardware within the terminal's data center	Air-gapped environments, strict data sovereignty, low-latency gate operations
Cloud	Turqoa runs on managed cloud infrastructure (AWS, Azure, GCP)	Multi-terminal deployments, elastic scaling, minimal on-site IT
Hybrid	AI inference runs on-premise for real-time processing; management plane runs in the cloud	Balance of latency requirements and operational flexibility

Architecture Overview

On-Premise Architecture

┌──────────────────── Terminal Data Center ────────────────────┐
│                                                              │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌────────────┐  │
│  │ Gate      │  │ AI       │  │ Decision │  │ Command    │  │
│  │ Controller│  │ Inference│  │ Engine   │  │ Center     │  │
│  │ (Edge)   │  │ (GPU)    │  │          │  │ (Web UI)   │  │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘  └────┬───────┘  │
│       └──────────────┴─────────────┴─────────────┘          │
│                        │                                     │
│               ┌────────▼────────┐                            │
│               │   PostgreSQL    │                            │
│               │   + Kafka       │                            │
│               └─────────────────┘                            │
└──────────────────────────────────────────────────────────────┘

Cloud Architecture

┌──────────── Cloud Region ────────────┐
│  ┌─────────┐  ┌─────────┐           │
│  │ K8s     │  │ Managed │           │
│  │ Cluster │  │ DB/MQ   │           │
│  └────┬────┘  └────┬────┘           │
│       └─────┬──────┘                │
│             │                       │
└─────────────┼───────────────────────┘
              │ VPN / Direct Connect
┌─────────────┼──── Terminal ────┐
│     ┌───────▼──────┐           │
│     │ Edge Gateway │           │
│     │ (Camera +    │           │
│     │  TOS proxy)  │           │
│     └──────────────┘           │
└────────────────────────────────┘

Hybrid Architecture

┌────────── Cloud ──────────┐     ┌──────── Terminal ────────────┐
│  ┌─────────────────────┐  │     │  ┌──────────┐  ┌──────────┐ │
│  │ Management Plane    │  │     │  │ AI       │  │ Gate     │ │
│  │ - User mgmt        │  │ VPN │  │ Inference│  │ Controller│ │
│  │ - Config sync       │◄─┼─────┼─►│ (GPU)    │  │ (Edge)   │ │
│  │ - Audit aggregation │  │     │  └──────────┘  └──────────┘ │
│  │ - Analytics         │  │     │  ┌──────────┐               │
│  └─────────────────────┘  │     │  │ Local DB │               │
└───────────────────────────┘     │  │ + Cache  │               │
                                  │  └──────────┘               │
                                  └─────────────────────────────┘

Decision Guide

Use the following criteria to select your deployment model:

Criteria	On-Premise	Cloud	Hybrid
Data stays on-site	Yes	No	Partial
Latency < 100ms (camera to decision)	Yes	Depends on connectivity	Yes
No internet dependency	Yes	No	Partial
Multi-site centralized management	Manual	Yes	Yes
Auto-scaling	No	Yes	Partial
Operational complexity	High	Low	Medium
Capital expenditure	High	Low	Medium
Recurring cost	Low	Medium-High	Medium

Recommendation: Most greenfield deployments should start with Hybrid. It provides real-time gate performance through on-premise inference while offloading management, analytics, and multi-site orchestration to the cloud.

Next Steps

On-Premise Deployment --- Hardware setup and installation
Cloud Deployment --- Cloud provider configuration and IaC templates
Requirements --- Detailed hardware, network, and storage specifications