Social graph storage in HLD - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Architecture

intermediate

Designing a scalable social graph storage system

You need to design a system to store and query a social graph with billions of users and their connections. Which architecture best supports fast retrieval of a user's friends and mutual friends?

A Use a relational database with tables for users and friendships, indexed by user IDs.

B Use a key-value store where each user ID maps to a list of friend IDs.

C Use a graph database optimized for traversals, storing users as nodes and friendships as edges.

D Use a document database storing each user document with embedded friend lists.

💡 Hint

Think about which storage type naturally represents relationships and supports fast graph queries.

Graph databases are designed to efficiently store and traverse relationships, making them ideal for social graphs where queries like mutual friends are common.

❓ scaling

intermediate

Scaling social graph storage for high write throughput

Your social graph system must handle millions of new friendship connections per minute. Which approach best supports this write load while maintaining query performance?

A Use a single master database to serialize all writes and replicate to read replicas.

B Partition the graph by user ID ranges and distribute partitions across multiple servers.

C Store all friendships in a single large cache to speed up writes.

D Batch all writes and apply them once per hour to reduce load.

💡 Hint

Consider how to distribute data to avoid bottlenecks and allow parallel writes.

Partitioning the graph by user ID distributes write load across servers, enabling high throughput and maintaining query speed by limiting data scope per server.

❓ tradeoff

advanced

Tradeoffs between consistency and availability in social graph storage

In a distributed social graph system, which choice best describes the tradeoff when prioritizing availability over consistency?

A Users may see stale friend data during network partitions but the system remains available.

B Users always see the most up-to-date friend list but the system may be unavailable during network issues.

C The system rejects writes during network partitions to maintain consistency.

D The system duplicates data to all nodes synchronously to ensure consistency.

💡 Hint

Think about the CAP theorem and what happens when availability is prioritized.

Prioritizing availability means the system continues to operate during network issues but may serve stale or inconsistent data temporarily.

🧠 Conceptual

advanced

Choosing data models for social graph storage

Which data model best supports efficient queries for "friends of friends" in a social graph?

A Relational tables with join queries on user and friendship tables.

B Key-value store with user IDs as keys and friend lists as values.

C Document store with embedded friend lists inside user documents.

D Graph model with nodes representing users and edges representing friendships.

💡 Hint

Consider which model naturally expresses multi-level relationships.

The graph model allows direct traversal of edges to find friends of friends efficiently without expensive joins or lookups.

❓ estimation

expert

Estimating storage requirements for a social graph

Estimate the storage needed to store a social graph with 1 billion users, each having an average of 200 friends. Assume each user ID and friend ID requires 8 bytes, and each friendship is stored bidirectionally.

A Approximately 3.2 terabytes

B Approximately 1.6 terabytes

C Approximately 6.4 terabytes

D Approximately 12.8 terabytes

💡 Hint

Calculate total friendships, multiply by bytes per friendship, and consider bidirectional storage.

1 billion users × 200 friends = 200 billion directed edges (bidirectional storage). Each edge stores two 8-byte IDs = 16 bytes. Total: 200B × 16 bytes = 3.2 TB.