Database Detective Game

Investigate IoT data clues and choose the database model that best fits the workload.

animation
database
data-management
game
interactive
learning
Learner-ready database detective game with case presets, Play/Step/Reset controls, workload clues, database suspect cards, scoring, CAP tradeoff notes, and mobile-ready learning support.
Interactive case lab IoT data architecture Database selection

Database Detective: read the IoT clues before you pick the database.

Step through each case file, inspect the workload evidence, shortlist the database suspects, and issue a verdict. The goal is not to memorize product names. It is to match the dominant data shape and query pattern to the model that handles it cleanly.

Case 1Current investigation
IntakeActive stage
0/0Closed cases
Pick a suspectCurrent verdict
1. Follow the query.The question users ask most often usually matters more than the database brand.
2. Size the write path.Sensor count, update rate, and payload size reveal whether history, latest state, or relationships dominate.
3. Check the promise.Transactions, availability, and partition behavior should match the risk of being stale or wrong.

Case file

Read the briefing, then use Play or Step to walk through the evidence board.

Beginner

Loading case file...

Device eventsSensors create evidence.
Write pathRate and payload set pressure.
Data shapeShape exposes the model.
Query lensThe question narrows the field.
VerdictPick the best fit.
clue
0 writes/s Write rate estimate 0 GB/day Raw daily ingest estimate using decimal KB and GB.

Detective Controls

Run the investigation, change cases, then check your verdict.

No verdict yet

Select a database suspect, then check the verdict.

Hint: start with the most frequent query, then use the warning signs to rule out weaker fits.

Database Suspects

Pick one model. The fit bars are shown after the evidence stage so learners can compare before answering.

UnknownBest database for this case
Review CAP noteConsistency and availability lens
Dominant clueWhat drives the decision

Suspect Fit Ranking

Scores are teaching weights for this case. They are not benchmark claims or vendor recommendations.

Database selection cheat sheet
  • Time-series: timestamped telemetry, append-heavy writes, time-window queries, downsampling, and retention policies.
  • Key-value: latest state, cache, session, feature flag, or sub-10 ms lookup where the key is already known.
  • Document: flexible nested records, varied device models, and whole-document reads without many cross-document joins.
  • Relational: ACID transactions, constraints, audits, joins, billing, inventory, and structured compliance reports.
  • Graph: multi-hop relationship traversal such as network topology, dependency impact, route reachability, and downstream effects.
CAP theorem guardrails

CAP applies to distributed systems during a network partition. In that moment, the system must favor consistency or availability. CA is only realistic when partitions are outside the model, such as a single-node system or a tightly controlled non-distributed deployment.

  • Use a CP leaning when wrong data is worse than temporary rejection, such as money, inventory, or safety interlocks.
  • Use an AP leaning when stale data is acceptable for a short time, such as dashboards, caches, or local read-mostly state.
  • Do not assume a database category always fixes CAP. Deployment topology and configuration matter.
Common detective mistakes
  1. Choosing time-series only because data has timestamps. If the main query is a join or transaction, timestamps are not enough.
  2. Choosing key-value for high speed when history and time-window analysis are required.
  3. Choosing document for flexibility when the real problem is relationship traversal.
  4. Choosing graph because devices form a network, even though the workload only stores periodic readings.
How the workload math is estimated

The game estimates raw ingest as devices x events per minute per device / 60 writes per second. Raw daily data uses writes per second x 86400 x payload KB / 1,000,000 GB/day. Real deployments add indexes, compression, replication, metadata, and retention rules.

IoT Database Selection Tool Time-Series Databases Data Storage and Databases Query Performance Analyzer