Database architecture
On this page 9 topics
Overview
Topics
- Managed SQL
- NoSQL
- caching
- replication
- DR
Learn first
3.5 h of practice (estimate)
Learning outcomes
- Select managed relational, NoSQL, cache, and search services per workload.
- Design replication, read scaling, and failover for each data store.
- Define backup, point-in-time recovery, and disaster recovery for data stores.
- Plan data migration from on-premises databases.
Lessons
Lab and exercises
Lab
ARC-016-LAB1 · Larkspur data platform
Outcomes: 1–4 · Time: ~3.5 h
- List Larkspur data workloads and their access patterns and consistency needs.
- Choose a data store per workload and name the source of truth.
- Design replication, read scaling, and failover per store.
- Define backup, PITR window, RPO, and RTO per store and compare with requirements.
- Write the migration approach for the current web shop database (method, cutover plan, rollback).
Acceptance: each store's RPO/RTO meets requirements or the gap is flagged.
Exercises (case-based; six exercises)
| ID | Outcome | Task |
|---|---|---|
| ARC-016-EX01 | 1 | Choose a store for carts, orders, product search, and a catalog read cache; justify each. |
| ARC-016-EX02 | 1 | List what Larkspur still owns after adopting a managed relational database. |
| ARC-016-EX03 | 2 | The reporting queries slow the order database. Propose a design and its staleness trade-off. |
| ARC-016-EX04 | 2 | Explain the difference between a multi-AZ standby and a read replica, with an example failure of each. |
| ARC-016-EX05 | 3 | A developer deleted rows at 14:05. Describe how to recover and what RPO the design gives. |
| ARC-016-EX06 | 4 | Outline a migration plan for the web shop database with 200 GB, a nightly low-traffic window, and a four-hour downtime limit. |
Show solutions
- EX01 Carts: key-value/document (simple, fast); orders: relational (transactions); search: managed search fed from the catalog; cache: in-memory cache, source of truth stays relational.
- EX02 Schema, indexes and queries, sizing, access and encryption, monitoring, restore testing, upgrade planning, cost.
- EX03 Add a read replica for reporting; accept replication lag of seconds to minutes; reports must tolerate stale data.
- EX04 Standby: survives zone failure with automatic failover, no extra read capacity. Replica: serves reads, may lag, promotion is a manual or scripted step.
- EX05 PITR to 14:04 into a new instance, copy back the rows, or switch with care; RPO is minutes, bounded by the log retention and backup configuration.
- EX06 Replicate continuously from on-premises, validate, rehearse, freeze writes in the window, final sync, switch connection, keep the old database read-only as the rollback; downtime minutes, within four hours.
Workplace evidence: a data platform design you led to production.
Outcome coverage: all outcomes covered by exercises and the lab.
Quiz
Outcomes: O1 select stores · O2 replication and failover · O3 backup, PITR, DR · O4 migration.
A self-check for you: nothing is graded, saved or sent anywhere. Choose an answer, then check it. You can try again as often as you like.
Practical part
Design review of the ARC-016-LAB1 data platform plus exercises ARC-016-EX01 to EX06. Pass: quiz at least 8/10, at least five of six exercises correct, and each store's RPO/RTO either meeting requirements or flagged as a gap.
Common mistakes
One database for everything; assuming managed means automatic DR.
Assessment
Design review.
Workplace evidence
A data platform design you led to production.
Career relevance
Data architecture (ARC-023).