consistencyreplicationavailabilitydistributed-storage
Quorum Consistency
Use read/write quorum sizes to balance consistency, availability, and latency in replicated stores.
Definition
Quorum consistency requires operations to be acknowledged by a subset of replicas such that read and write quorums overlap.
When To Use
- Replicated data stores needing tunable consistency.
- Systems that must survive replica/node failures without total outage.
- Workloads that can trade strict linearizability for lower latency or availability.
When Not To Use
- Strongly serializable workflows without compensating mechanisms.
- Very low-latency paths where quorum fanout is too expensive.
- Scenarios with severe clock/network instability and weak repair.
Tradeoffs
- Increases availability under failures, but can return stale reads under some settings.
- Improves durability confidence, but increases tail latency due to multi-replica coordination.
- Allows tunable R/W policies, at the cost of repair complexity.
Common Failure Modes
- Replica lag causes stale reads despite successful quorum.
- Network partitions produce split-brain writes with conflict resolution debt.
- Anti-entropy backlog grows and consistency drifts across regions.
Interview Framing
Use this structure when the interviewer asks for this pattern explicitly.
State your R/W/N values, overlap guarantees, read-repair strategy, and behavior under partition.
Related Project Deep Dives
Distributed Key-Value Store
Design a strongly reliable key-value store with partitioning, replication, quorum reads/writes, and predictable low-latency access.
Distributed Cache System
Design a distributed cache system like Redis or Memcached that handles millions of requests per second with sub-millisecond latency, high availability, and intelligent eviction policies.
Related Concepts
Consistent Hashing
Distribute keys across nodes while minimizing remapped keys during node add/remove events.
Leader Election
Select a single coordinator for shared work while preserving failover safety.
Exactly-Once Processing (Practical)
Achieve effective exactly-once outcomes via idempotency, transactions, and dedup rather than magic guarantees.