HLDArena
HLD-406·Storage & Caching
Expert · 2,000 pts

Partitioned Quorum Key-Value Store (Dynamo / Cassandra)

Design a planet-scale consistent-hashed Key-Value storage engine supporting `R + W > N` quorum consistency, hot-partition splitting/salting, and sub-second node crash recovery.

Max p99
≤ 75 ms
Availability
≥ 99.5%
Budget
≤ $11,600

Functional Requirements

  • FR-1PUT /kv/{partition_key} — Commits versioned mutation across `W` quorum replicas with WAL durability.
  • FR-2GET /kv/{partition_key} — Executes `R` quorum read/cache lookup returning latest committed version.
  • FR-3DELETE /kv/{partition_key} — Releases canceled or expired leases/holds atomically and immediately refreshes cached availability.
  • FR-4Must guarantee linearizable/quorum read-after-write and zero data loss across storage node crashes at 240k RPS.

Workload Profile

Baseline Cluster Load82,000 Read RPS / 14,000 Quorum Write RPS
Hot-Partition Storm210,000 Read RPS / 34,000 Quorum Write RPS
Active Working Set56 GB hot LSM memtables & block/row caches
Consistency GateStrict R+W>N quorum consistency + RPO = 0 failover

Consistency Model

Strict quorum (`R + W > N`) + atomic version CAS so concurrent writers never clobber or serve stale values.

Relevant practice for

Click an item in Inventory above, or press B to add a component, double-click to name, C to connect.

Design Write-up

Write under ### Architecture for Diagram to take that up directly

Compulsory section · if not written yet, Diagram reads from the start of your write-up.
Architecture: required·FRs mentioned: 0/4·0 boxes · 0 edges
0 words
Hidden Cases #4–#7 run automatically on Submit
Case 1: Smoke & API ContractInput Workload
Read RPS
2,000 rps
Write RPS
400 rps
Key Distribution
Uniform (Low Skew)
Chaos Fault
None (Healthy Cluster)
Expected Pass CriteriaASSERTIONS
• Tail Latency: p99 ≤ 150 ms
• Availability: ≥ 99.5%
Valid end-to-end Client → Compute → Storage topology