Architect a scalable Feed Generation Engine

Question

Design a scalable feed generation system that handles millions of requests. Discuss trade-offs in consistency, availability, and performance.

Codemia · Accepted Answer

# Functional Requirements
1. **User Feed Generation**: The system should generate personalized feeds for users based on their interaction and preferences across Atlassian products like Jira, Confluence, and Trello.
2. **Real-Time Updates**: The feed should reflect real-time updates as users perform actions within the applications (e.g., comments, task updates).
3. **Feed Customization**: Users should be able to customize what types of updates they want to see in their feed.
4. **Notifications**: The system should support sending notifications related to feed updates via email or in-app notifications.

# Non-Functional Requirements
1. **Scalability**: The system should handle at least 10 million requests per day, with the ability to scale horizontally.
2. **Availability**: The system should maintain 99.9% uptime.
3. **Performance**: The feed generation should occur in under 200 milliseconds for each request.
4. **Security**: The system must authenticate users securely and prevent unauthorized access to feeds.

Assuming an average of 10 million requests per day:
- **Requests per second (RPS)**: 10,000,000 requests / (24 * 60 * 60) = ~115 RPS
- To account for spikes, we might consider scaling to handle 300 RPS during peak times.
- Assuming each feed generation requires a few database reads and writes (let's estimate 3 for reads and 1 for writes), our database should handle around 1,200 operations per second (300 * 4).
- Considering a 2x buffer for efficiency and failover, we should design our database layer to handle at least 2,400 operations per second.

The architecture consists of the following components:
1. **API Gateway**: Acts as the entry point for all requests, routing them to appropriate microservices. Technologies: AWS API Gateway or Kong.
2. **Feed Service**: This microservice generates and serves user feeds. Technologies: Node.js or Go for performance.
3. **Data Store**: A NoSQL database like DynamoDB for fast reads/writes and scalability.
4. **Message Queue**: Use Kafka or RabbitMQ for asynchronous processing of feed updates.
5. **Cache Layer**: Redis for caching frequently accessed feeds to reduce database load.
6. **Monitoring and Logging**: Use tools like Prometheus and Grafana for monitoring, and ELK stack for logging.

The data model will consist of the following collections/tables:
1. **Users**: (user_id, username, email, preferences)
2. **Feeds**: (feed_id, user_id, created_at, content)
3. **Feed_Updates**: (update_id, feed_id, update_type, update_content, timestamp)

This schema allows us to efficiently fetch a user's feed based on their preferences and the latest feed updates. Each user's feed can be generated by querying the Feed_Updates table and filtering based on user preferences.

1. **Consistency vs. Availability**: Choosing eventual consistency for feed updates allows us to provide high availability and performance but may lead to slightly stale data in user feeds.
2. **Caching vs. Freshness**: Using Redis as a cache improves performance but requires a strategy for cache invalidation to ensure users see the latest updates.
3. **Synchronous vs. Asynchronous Processing**: While synchronous processing could ensure immediate feed updates, it would add latency. Asynchronous processing with message queues allows for better load management but introduces a delay in updates appearing in feeds.
4. **Cost vs. Performance**: Using managed services for databases and caching may incur higher costs but can simplify scaling and reduce operational overhead.

Architect a scalable Feed Generation Engine

Atlassian

What the Interviewer Expects

Key Topics to Cover

How to Approach This

Possible Follow-up Questions

Practice a Similar Problem on Codemia

Sample Answer

Requirements

Functional Requirements

Capacity Estimation

Submit Your Answer

Atlassian Software Engineer Interview Guide

Related Questions

Design a high-throughput Inventory Management System

Design a low-latency Rate Limiting System

Design a fault-tolerant Payment System

Design a fault-tolerant Messaging System

Design Walmart Product Search