Course contentsShow
← System Design
Lesson 315 of 1,9199. SQL vs NoSQL TradeoffsFree lesson

Basically Available: Prioritizing Uptime

How BASE systems favor availability over consistency, allowing partial responses during failures.

Basically Available: Prioritizing Uptime

What You'll Learn

You'll understand how BASE systems sacrifice strict consistency to keep services running even when parts of the system fail.

The "Basically Available" Philosophy

In the BASE model (Basically Available, Soft state, Eventually consistent), "Basically Available" means the system prioritizes staying operational above all else. Unlike ACID systems that might reject requests or halt operations to maintain perfect consistency, BASE systems choose to respond—even if that response is incomplete or slightly stale.

How It Works in Practice

When a BASE-compliant system encounters failures (network partitions, node crashes, overload), it doesn't stop serving requests. Instead, it:

  • Returns partial data from available nodes
  • Serves slightly stale data from a cache or replica
  • Degrades functionality gracefully rather than failing completely
  • Accepts writes that will be reconciled later

Real-World Analogy

Imagine a grocery store during a power outage. An ACID approach would close entirely until power returns (maintaining perfect accuracy). A "Basically Available" approach keeps operating with flashlights and manual price lookups—service continues, though it's slower and might have minor pricing errors that get corrected later.

Trade-offs

You gain resilience and uptime at the cost of temporary inconsistency. Users might see different product inventory counts on different servers, or recent updates might not appear immediately—but the service never goes completely dark.

This approach is ideal for systems where availability matters more than instant accuracy: social media feeds, product catalogs, recommendation engines, and shopping carts.

Key Takeaway: Basically Available systems choose to stay online and serve requests during failures, accepting temporary inconsistencies rather than system downtime—prioritizing user access over perfect data accuracy.