Deployment

Deployment Guide#

Choose the deployment strategy that best fits your requirements.

Deployment Modes#

ncps supports two primary deployment modes:

Single-Instance Deployment#

One ncps server handling all cache requests.

Best for:

Development and testing
Small to medium teams (1-100+ users)
Single-location deployments
Simpler operations

Characteristics:

One server
Local locks (no Redis)
Local or S3 storage
SQLite, PostgreSQL, or MySQL database
Easier to set up and maintain

Learn more →

High Availability Deployment#

Multiple ncps instances for redundancy and scalability.

Best for:

Production environments
Large teams (100+ users)
Business-critical infrastructure
Geographic distribution
Zero-downtime requirements

Characteristics:

2+ servers
Redis distributed locking
S3 storage (required)
PostgreSQL or MySQL database (required, NOT SQLite)
Load balancer
More complex setup

Learn more →

Quick Comparison#

Aspect	Single-Instance	High Availability
Servers	1	2+
Locking	Local (in-process)	Redis distributed locks
Storage	Local or S3	S3 (required)
Database	SQLite, PostgreSQL, or MySQL	PostgreSQL or MySQL (NOT SQLite)
Load Balancer	Not needed	Required
Redundancy	None	Full
Complexity	Simple	Moderate
Zero Downtime	No	Yes
Scalability	Limited to one server	Horizontal scaling
Cost	Lower	Higher

Decision Tree#

Start
  │
  ├─ Need zero-downtime updates? ──Yes──> High Availability
  ├─ Need geographic distribution? ──Yes──> High Availability
  ├─ Team > 100 users? ──Yes──> High Availability
  ├─ Mission-critical service? ──Yes──> High Availability
  │
  └─ Otherwise ──> Single-Instance

Documentation#

Single-Instance Deployment - Deploy one ncps server
High Availability Deployment - Deploy multiple instances with HA
Distributed Locking - Deep dive into Redis locking for HA

Prerequisites by Mode#

Single-Instance Prerequisites#

Minimum:

Server or VM (2+ CPU cores, 4GB+ RAM recommended)
Storage (50GB-1TB depending on usage)
Network connectivity to upstream caches

Optional:

S3-compatible storage (for cloud-native or future HA)
PostgreSQL/MySQL (for better performance than SQLite)

High Availability Prerequisites#

Required:

2+ servers (3+ recommended for better availability)
Redis server (single instance or cluster)
S3-compatible storage (AWS S3, MinIO, etc.)
PostgreSQL or MySQL database
Load balancer (nginx, HAProxy, cloud LB)

Optional:

Monitoring and alerting (Prometheus, Grafana)
Centralized logging (ELK, Loki)

Getting Started#

Choose deployment mode based on your requirements
Review prerequisites for your chosen mode
Follow installation guide:
Configure according to your mode:
- Single Instance
- High Availability
Verify deployment and test
Set up monitoring (recommended)

Migration Path#

From Single-Instance to HA#

Common migration path as your needs grow:

Start: Single instance with local storage and SQLite
Scale up: Move to PostgreSQL for better performance
Cloud-ready: Migrate to S3 storage
High Availability: Add Redis and additional instances

Each step is incremental and can be done independently.

See High Availability Guide for detailed migration steps.

Common Deployment Patterns#

Pattern 1: Development Setup#

Developer Workstation
└── ncps (Docker)
    ├── Local storage
    └── SQLite

Pattern 2: Small Team#

Shared Server
└── ncps (systemd)
    ├── Local NFS storage
    └── SQLite or PostgreSQL

Pattern 3: Cloud Production (Single)#

Cloud VM
└── ncps (Docker/Kubernetes)
    ├── S3 storage
    └── Managed PostgreSQL (RDS, etc.)

Pattern 4: High Availability#

Load Balancer
├── ncps Instance 1
├── ncps Instance 2
└── ncps Instance 3
    ├── Shared S3 storage
    ├── Shared PostgreSQL
    └── Shared Redis

Next Steps#

Choose and follow deployment guide
Configure clients to use your cache
Set up monitoring for production
Review operations guides for maintenance