User
Aurora DSQLStable relational record with session and score history.
Blueprint Canvas
Global Trivia Game
1,000,000 users, 50,000 peak concurrent, US + EU traffic. Score updates require strong settlement, leaderboards can be near real-time, game events are append-only.
Entities
GameSession
HybridAurora DSQL owns metadata; DynamoDB stores live event pressure.
Question
Aurora DSQLAuditable relational content with schema versioning.
AnswerEvent
DynamoDBBurst writes and replay by run/session.
Score
HybridTransactional settlement plus event replay.
LeaderboardEntry
DynamoDBHigh-read, write-sharded rank access.
Region
Aurora DSQLReference data for regional data domains.
Generated flow
AP-01
Join live game session
Attach a user to an active game session and resolve regional routing.
AP-02
Submit answer event
Append answer telemetry when a player taps an answer during a live round.
AP-03
Update score atomically
Settle score after validating answer correctness and anti-cheat signals.
AP-04
Get top 100 leaderboard by region
Read top ranked players for a regional board with near-real-time freshness.
AP-05
Get top 100 global leaderboard
Read a global leaderboard view while keeping writes distributed.
AP-06
Get user session history by time range
Query historical sessions for support, profile pages, and audit review.
AP-07
Replay disputed scoring event
Reconstruct score history from immutable answer events during support review.
AP-08
Get user profile by userId
Load player profile, preferences, and account region before a session.
Database Mapping
Aurora DSQL
- Project metadata and B2B workspace records
- Access pattern registry and recommendation versions
- Schema versions, simulation summaries, reports
DynamoDB
- Answer events and representative workload telemetry
- Leaderboard read/write samples and hot-key metrics
- Run timeline, event replay, risk-indexed samples
Access Patterns
Each row is persisted and feeds the recommendation version.
| Pattern | Frequency | Consistency | Latency | Database | Peak QPS |
|---|---|---|---|---|---|
Join live game session Session membership is relational, but live events should stay decoupled. Impact: QPS 11,000 -> 11,000; hot risk high; shards 28. | 100 ms | Hybrid | 11,000 | ||
Submit answer event The wrong key design concentrates writes on a single game partition. Impact: QPS 8,000 -> 8,000; hot risk high; shards 20. | 35 ms | DynamoDB | 8,000 | ||
Update score atomically Score settlement needs transactional facts plus event replay. Impact: QPS 4,000 -> 4,000; hot risk moderate; shards 10. | 120 ms | Hybrid | 4,000 | ||
Get top 100 leaderboard by region Regional buckets avoid a global leaderboard hot key. Impact: QPS 25,000 -> 25,000; hot risk critical; shards 63. | 45 ms | DynamoDB | 25,000 | ||
Get top 100 global leaderboard A global partition must be materialized from regional shards. Impact: QPS 19,000 -> 19,000; hot risk critical; shards 48. | 60 ms | DynamoDB | 19,000 | ||
Get user session history by time range Time-range history and support queries are easier to govern in SQL. Impact: QPS 2,000 -> 2,000; hot risk moderate; shards 5. | 150 ms | Aurora DSQL | 2,000 | ||
Replay disputed scoring event Event replay needs DynamoDB history plus Aurora DSQL authoritative facts. Impact: QPS 200 -> 200; hot risk moderate; shards 4. | 300 ms | Hybrid | 200 | ||
Get user profile by userId Relational identity data benefits from SQL constraints and joins. Impact: QPS 15,000 -> 15,000; hot risk critical; shards 38. | 80 ms | Aurora DSQL | 15,000 |