System Design: Notification Service

Introduction

• Objective: Design a notification service for web services.
• Scenarios:
  • Card transaction exceeds a limit.
  • Service monitoring detects faults.
• Publisher: Produces messages.
• Subscribers: Receives messages.
• Challenges of synchronous communication:
  • Scalability issues.
  • Hard to extend.

Proposed Solution

• Asynchronous System: Allows many publishers and subscribers.
• Functional Requirements:
  • Create a topic.
  • Publish a message to a topic.
  • Subscribe to receive messages.
• Non-Functional Requirements:
  • Scalability, maintainability, testability.
  • High availability, fast delivery, durability.

Architecture Overview

• Load Balancer: Distributes requests.
• Frontend Service: Initial request processing.
• Metadata Service: Caching layer, separation of concerns.
• Message Storage: Temporary storage for delivery.
• Sender Component: Retrieves from storage and sends to subscribers.

Detailed Component Analysis

Frontend Service

• Responsibilities: Validation, authentication, SSL termination, encryption, caching, throttling, logging.
• Reverse Proxy: Handles SSL, compression, response handling.
• Local Cache: Reduces calls to metadata service.
• Logging and Metrics: Service health, performance monitoring.

Metadata Service

• Role: Stores topics and subscriptions.
• Distributed Cache: Consistent hashing ring, no coordinator or with coordinator.

Temporary Storage Service

• Purpose: Fast, available, scalable message storage.
• Options:
  • SQL vs NoSQL (e.g., Apache Cassandra, Amazon DynamoDB)
  • In-memory (e.g., Redis)
  • Message queues (e.g., Apache Kafka, Amazon Kinesis)

Sender Component

• Message Retrieval: Pool of threads, dynamic scaling with semaphores.
• Metadata Calls: Retrieves subscriber info.
• Task Processing: Splits delivery tasks for parallel processing.

Additional Considerations

• Spam Prevention: Subscriber confirmation, deduplication.
• Retry Policies: Delivery retries, subscriber responsibility for duplicates.
• Message Order: No guaranteed delivery order.
• Security: Authentication, encryption, SSL, monitoring.
• Monitoring and Metrics: Track message state and system health.

Evaluation

• Scalable: Horizontally and vertically scalable components.
• High Availability: No single point of failure, deployed across data centers.
• Performance: Fast processing, efficient task management.
• Durability: Redundant data storage and message retries.

Conclusion

• Comprehensive design for a notification service.
• Addressed both functional and non-functional requirements.
• Encouraged to explore more on monitoring and specific storage solutions.