Course contentsShow
← System Design
Lesson 583 of 1,91916. Distributed TransactionsFree lesson

Alternative: Best Effort with Eventual Consistency

Accepting temporary inconsistency and relying on reconciliation processes over strict atomicity.

Alternative: Best Effort with Eventual Consistency

What You'll Learn

You'll discover how systems can accept temporary inconsistency and use background reconciliation instead of strict distributed transactions to achieve better availability and performance.

The Core Idea

Instead of forcing all systems to agree atomically (like with Two-Phase Commit), best effort with eventual consistency lets operations succeed immediately, accepts that data might be temporarily out of sync, and relies on background processes to fix inconsistencies later.

Think of it like splitting a restaurant bill: instead of everyone calculating their exact share before anyone pays (blocking and slow), one person pays upfront, and you settle up via Venmo later. The payment happens immediately, even if the final accounting takes time.

How It Works

When an operation spans multiple systems:

  1. Execute locally first - Each system commits its part independently
  2. Accept temporary inconsistency - Data might be out of sync for seconds, minutes, or longer
  3. Reconcile asynchronously - Background jobs detect and fix mismatches
  4. Handle conflicts gracefully - Use application logic or CRDTs to resolve discrepancies

For example, when you place an online order, the order service might succeed immediately while the inventory system updates separately. If inventory is actually out of stock, a reconciliation process catches this later and sends you a "sorry, refunded" email rather than blocking your checkout.

Trade-offs

Benefits: High availability (no blocking), better performance, partition tolerance

Costs: Temporary inconsistency, complex reconciliation logic, potential user-facing corrections

This approach works well for systems where availability matters more than instant consistency—social media posts, shopping carts, analytics pipelines.

Key Takeaway: Best effort with eventual consistency trades strict atomic guarantees for availability and performance, accepting temporary inconsistency that background processes fix over time.