Build a distributed Ride Matching Pipeline

Question

Design a distributed ride matching system that handles millions of requests. Discuss trade-offs in consistency, availability, and performance.

Codemia · Accepted Answer

### Functional Requirements
- **Real-time Ride Matching**: Match drivers and riders in real-time based on proximity, destination, and preferences.
- **User Profiles**: Manage profiles for riders and drivers, including ratings and ride history.
- **Ride History**: Store and retrieve past ride details for analysis and recommendations.
- **Feedback Mechanism**: Allow users to provide feedback on rides, which will be used for model retraining.

### Non-Functional Requirements
- **Scalability**: System should handle up to 10 million requests per day, with the ability to scale horizontally.
- **High Availability**: The system must be available 99.9% of the time to ensure user satisfaction.
- **Low Latency**: Ride matching should occur in under 2 seconds to provide a good user experience.
- **Data Consistency**: Ensure that ride status updates are consistently reflected across all services.

To estimate capacity:
- **Daily Requests**: Assume 10 million ride requests per day, translating to approximately 115 requests per second (10 million / 86400 seconds).
- **Data Storage**: Each ride request may generate about 1 KB of data (user info, location, timestamps), leading to roughly 10 GB of new data daily.
- **Feature Store**: With features derived from user profiles and ride history, we anticipate needing around 100 GB of storage for features, growing as we include more attributes.
- **Model Training**: Assume a model retrains weekly; we’ll need 1 TB of historical data for effective training, which requires careful data retention and archiving strategies.

### Component Diagram
- **API Gateway**: Handles incoming requests and routes them to appropriate services.
- **Ride Matching Service**: Core service utilizing a machine learning model to match riders with drivers based on features.
- **Feature Store**: A centralized repository (using technologies like **Apache Iceberg** or **Featurestore** by Tecton) for storing and retrieving features.
- **Model Serving Infrastructure**: Deployed using **TensorFlow Serving** or **Seldon** for low-latency predictions.
- **Data Pipeline**: Built with **Apache Kafka** for real-time data ingestion and **Apache Spark** for batch processing.
- **Monitoring and Logging**: Tools like **Prometheus** and **Grafana** for real-time metrics and **ELK Stack** for logging.

### Schema Design
- **User Table**: (user_id, name, phone_number, rating, ride_history)
- **Driver Table**: (driver_id, name, car_details, rating, availability_status)
- **Ride Request Table**: (request_id, user_id, driver_id, pickup_location, dropoff_location, request_time, status)
- **Feedback Table**: (feedback_id, request_id, user_id, rating, comments, feedback_time)

### Access Patterns
- Frequent reads for user profiles and ride history for matching.
- Writes for ride requests and updates based on availability and feedback.

### Architectural Decisions
- **Consistency vs. Availability**: Chose eventual consistency for ride status updates to ensure high availability during peak loads. This allows for fast ride confirmations even if some updates are not immediately reflected.
- **Cold Start Problem**: Implemented a hybrid approach with user-based collaborative filtering for new users while using popular routes for initial matching.
- **Model Freshness**: Adopted a canary deployment strategy for new models, allowing us to test new updates with a small percentage of traffic before full rollout.
- **Fairness and Bias Mitigation**: Introduced fairness metrics in the training process, adjusting weights in the loss function to ensure all user demographics are served equitably.

Build a distributed Ride Matching Pipeline

HRT

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

HRT 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