Skip to main content

Communication Service Architecture

Overview​

The Communication Service is built with Elixir/Phoenix and leverages OTP (Open Telecom Platform) for high concurrency and fault tolerance.

Architecture Diagram​

┌─────────────┐
│   Gateway   │
└──────┬──────┘
       │
       ▼
┌─────────────────────┐
│ Communication API   │
│  (Phoenix/GraphQL)  │
└──────┬──────────────┘
       │
       ├──► Oban (Background Jobs)
       │         │
       │         ├──► Email Worker
       │         ├──► SMS Worker
       │         └──► Push Worker
       │
       ├──► PostgreSQL (Templates, Tracking)
       │
       └──► External Providers
              ├──► SMTP Servers
              ├──► Twilio
              ├──► BulkSMS.com
              ├──► AWS SNS
              ├──► FCM
              └──► APNS

Components​

API Layer (Phoenix)​

  • GraphQL API - Primary interface using Absinthe
  • REST API - Alternative REST endpoints
  • WebSocket - Real-time delivery status updates

Background Processing (Oban)​

  • Email Worker - Handles email sending asynchronously
  • SMS Worker - Processes SMS requests with provider failover
  • Push Worker - Manages push notification delivery

Storage​

  • PostgreSQL - Templates, delivery tracking, configuration
  • Redis - Oban job queue and caching

Providers​

  • Email: SMTP with multiple provider support
  • SMS: Twilio, BulkSMS.com, AWS SNS (automatic failover)
  • Push: FCM (Android), APNS (iOS)

Design Patterns​

Supervisor Tree​

All workers run under supervision trees for fault tolerance. If a worker crashes, it's automatically restarted.

Retry Logic​

Failed deliveries are automatically retried with exponential backoff:

  • Initial delay: 1 second
  • Max delay: 60 seconds
  • Max retries: 5

Provider Failover​

For SMS, if the primary provider fails, the system automatically tries the next provider in the configured order.

Scalability​

  • Horizontal Scaling: Stateless design allows multiple instances
  • Concurrency: Elixir processes handle thousands of concurrent requests
  • Background Jobs: Heavy operations run asynchronously via Oban

Monitoring​

  • Prometheus Metrics - Request rates, error rates, delivery success
  • OpenTelemetry Tracing - Distributed tracing across providers
  • Health Checks - Database and Redis connectivity monitoring