import { Callout } from 'nextra/components'

# AWS Architecture

Suno's infrastructure runs primarily on AWS, with Modal for GPU workloads.

## Compute Services

### ECS (Elastic Container Service)

- Django API servers running on Fargate/EC2
- Celery worker pools for background jobs
- Auto-scaling based on CPU/memory

### Lambda

- Event processing (Kinesis consumers)
- Database monitoring and alerts
- Scheduled maintenance tasks

### Modal (Third-party)

- GPU-based ML inference
- Music and video generation
- Audio processing workloads

See [Modal Infrastructure](/onboarding/modal-infrastructure) for details.

## Storage Services

### S3 (Simple Storage Service)

- Generated audio files (MP3, WAV)
- Generated images and videos
- Model checkpoints and artifacts
- Analytics data (via Kinesis Firehose)

### EBS (Elastic Block Store)

- PostgreSQL database volumes
- Persistent storage for ECS tasks

### ElastiCache

- Redis/Valkey clusters for caching
- Multi-AZ deployment with automatic failover

## Networking

### CloudFront

- CDN for audio/video delivery
- Edge caching for static assets
- Custom domain support

### Application Load Balancer

- Distributes traffic to Django API servers
- Health checks and automatic failover
- HTTPS termination

### VPC (Virtual Private Cloud)

- Multi-region network setup
- Private subnets for databases
- Public subnets for load balancers

## Data Processing

### Kinesis

- Real-time event streaming
- User activity tracking
- Analytics pipeline

### Firehose

- Delivers events to S3
- Batching and compression
- Integration with analytics tools

### Glue

- ETL jobs for analytics
- Data transformations
- Schema discovery

## Monitoring & Observability

### CloudWatch

- AWS service logs and metrics
- Custom metric namespaces
- Alarms and notifications

### Datadog

- APM (Application Performance Monitoring)
- Custom metrics via StatsD
- Dashboard and alerting
- Log aggregation

### Sentry

- Error tracking and reporting
- Performance monitoring
- Release tracking
