Latency-Consistency Spectrum
1 min read
Map where your feature sits on the spectrum from low-latency/eventual to high-latency/strong.
Place each feature on the spectrum from eventual (fast) to strong (slow). Per-operation choices are more impressive than per-system choices.
How It Works
Every distributed system feature sits somewhere on the latency-consistency spectrum (see Consistency for the underlying models). Strong consistency costs a round-trip to the leader. Eventual consistency returns immediately from the nearest replica. Placing different features at different points on this spectrum is a high-signal answer — it shows per-operation trade-off reasoning.
Real-World Example
In a banking app: account balance after your own transfer -> read-your-writes. Transaction history -> bounded staleness with 5-second lag. ATM withdrawal authorization -> linearizable. Each feature has a different latency-consistency requirement.
Test Yourself
Scenario: A social app has two features that both update counters shown in the UI: posting a tweet (updates the user's tweet count) and liking a tweet (updates the tweet's like count). Pick a consistency model for each and justify.
Get notified when we launch
One email when the full practice product is live. No spam.