Course contentsShow
← System Design
Lesson 1550 of 1,91938. Real-World System: PastebinFree lesson

Object Storage for Paste Content

Storing large paste content in S3 or similar blob storage rather than in relational database.

Object Storage for Paste Content

You'll learn why large paste content belongs in blob storage like S3 instead of database tables.

When users paste code snippets, documents, or text, that content can range from a few bytes to several megabytes. Storing this directly in your relational database (like PostgreSQL or MySQL) creates serious problems at scale.

The Problem with Database Storage

Relational databases excel at structured queries, indexing, and transactional operations—but they're not optimized for storing large binary or text blobs. When you store paste content directly in database rows:

  • Table bloat: Large TEXT/BLOB columns make tables enormous, slowing down backups and maintenance
  • Memory pressure: Database buffer pools get filled with paste content instead of indexes and metadata
  • Expensive reads: Every paste retrieval loads megabytes through the database, wasting I/O bandwidth
  • Cost inefficiency: Database storage is typically 10-20× more expensive than object storage

The Object Storage Solution

Think of your database as a filing cabinet index and object storage (S3, Google Cloud Storage, Azure Blob) as a warehouse. The database stores lightweight metadata:

pastes table:
- paste_id (PK)
- user_id
- title
- created_at
- s3_key  ← points to actual content

The actual paste content lives in S3 with a key like pastes/2024/01/abc123.txt. When a user requests a paste, you:

  1. Query the database for metadata (fast, small)
  2. Fetch content from S3 using the stored key
  3. Return both to the user

Object storage gives you durability (11 nines), automatic scaling, CDN integration, and pennies-per-GB pricing.

Key Takeaway:
Store large paste content in object storage (S3) and keep only metadata in your relational database—this separates concerns, reduces costs, and scales reads independently.