Course contentsShow
← System Design
Lesson 494 of 1,91914. CAP TheoremFree lesson

AP Systems: Prioritizing Availability

Understanding systems that choose availability over consistency during network partitions and their tradeoffs.

AP Systems: Prioritizing Availability

What You'll Learn

You'll discover how some distributed systems choose to remain operational during network partitions, even if it means serving stale or inconsistent data.

The Core Idea

When a network partition splits your system, AP systems prioritize Availability over Consistency. This means all nodes continue accepting requests and serving responses, even though they can't communicate with each other to coordinate a consistent view of data.

Think of it like franchise coffee shops during a communications blackout. Each location keeps serving customers using whatever inventory data they last had, rather than shutting down until they can verify stock levels with headquarters. Some shops might sell items already sold out elsewhere, but everyone gets served.

How AP Systems Work

During a partition, AP systems:

  • Accept writes on all sides of the partition
  • Serve reads from local data, which may be outdated
  • Queue conflicting updates for later reconciliation
  • Prioritize uptime over correctness

Real-World Examples

  • DynamoDB (with eventual consistency settings)
  • Cassandra (at lower consistency levels)
  • DNS systems
  • Shopping cart services that never want to reject an "add to cart" operation

The Trade-offs

You gain:

  • Zero downtime during partitions
  • Low latency responses
  • Horizontal scalability

You accept:

  • Temporary inconsistencies between nodes
  • Conflict resolution complexity after partition heals
  • Risk of users seeing different data simultaneously

Key Takeaway: AP systems choose to stay available during network failures, accepting that data may be temporarily inconsistent and require reconciliation later—ideal when uptime matters more than perfect accuracy at every moment.