Design a Load Balancing for Figma

Question

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

Codemia · Accepted Answer

- **Functional Requirements:**
  - Distribute incoming API requests across multiple data centers to ensure low latency and high availability.
  - Support for geo-aware routing to direct users to the nearest data center based on their location.
  - Implement health checks and automatic failover to ensure requests are routed away from unhealthy instances.
  - Rate limiting to protect the backend services from overload and abuse.

- **Non-Functional Requirements:**
  - Must handle at least 5 million requests per second during peak usage times.
  - Latency should be under 100 milliseconds for 95% of requests.
  - Ensure data consistency across regions, with a focus on eventual consistency where appropriate.
  - High availability with an SLA of 99.99% uptime.
  - Secure API endpoints with OAuth 2.0 for authentication and authorization.

- Assuming Figma has 5 million active users, each generating an average of 1 request per second:
  - Total requests per second (RPS) = 5 million.
  - Considering peak traffic, we estimate up to 10 million RPS during high-demand periods.
  - Each request may require at least 100 ms to process and return a response.
  - To handle 10 million RPS effectively, we will need to provision sufficient load balancers, application servers, and databases across multiple regions.
  - With an average of 100 requests per second per server, we would need approximately 100,000 servers distributed globally (10 million RPS / 100 RPS per server).

- **Component Diagram:**
  - **Load Balancer (NGINX or HAProxy):** Distributes incoming requests based on geo-location and server health status.
  - **API Gateway (Kong):** Provides API management, rate limiting, and authentication.
  - **Application Servers (Node.js):** Handle the main application logic, providing scalability through horizontal scaling.
  - **Database (Cassandra):** For distributed data storage with eventual consistency, allowing for high write and read throughput.
  - **Cache (Redis):** To cache frequently accessed data and reduce load on the database.
  - **Monitoring (Prometheus + Grafana):** For real-time health checks, metrics collection, and alerting.
  - **Distributed Tracing (Jaeger):** For tracing requests and identifying bottlenecks in the system.

- **Schema Design:**
  - **User Table:**
    - `user_id (UUID) PRIMARY KEY`
    - `email (VARCHAR)`
    - `created_at (TIMESTAMP)`
    - `last_active (TIMESTAMP)`
  - **Project Table:**
    - `project_id (UUID) PRIMARY KEY`
    - `user_id (UUID) FOREIGN KEY`
    - `project_name (VARCHAR)`
    - `created_at (TIMESTAMP)`
  - **Collaborators Table:**
    - `project_id (UUID) FOREIGN KEY`
    - `user_id (UUID) FOREIGN KEY`
    - `role (ENUM)`
  - **Access Patterns:**
    - Frequent reads for user data and projects, necessitating caching layers.
    - Write-heavy during project creation and updates, managing eventual consistency across regions.

- **Consistency vs. Availability:**
  - Choosing eventual consistency for user-generated content to ensure availability and performance but requiring conflict resolution strategies.
  - Utilizing a quorum-based write strategy in Cassandra to balance consistency and availability.

- **Load Balancing Strategy:**
  - Geo-based routing for low latency but may lead to inconsistencies if data is not synchronized promptly across regions.
  - Sticky sessions may enhance performance but could reduce fault tolerance.

- **Operational Excellence:**
  - Implementing a phased rollout strategy with canary releases to minimize risks during deployment.
  - Regular chaos engineering drills to ensure resilience and quick recovery from failures.
  - Setting SLOs to monitor the performance and availability of the load balancer and backend services.

Design a Load Balancing for Figma

Figma

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

Capacity Estimation

Submit Your Answer

Figma 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