System Design — Staff

1Architecture at org scale

0/5▶

Staff rounds are about systems of systems: global scale, cross-team boundaries, and decisions that outlive any single service.

The big picture

Multi-region: active-passive vs active-active; data locality, latency, failover.

AWS Builders' Library.

Global consistency vs availability — geo-partitioning, consensus (Raft/Paxos).

Raft — visual.

Cell-based architecture to bound blast radius; shuffle sharding.

AWS — Cell-based.

Service boundaries & data ownership; when a monolith beats microservices (Conway).

Fowler — Monolith First.

Async, event-driven backbones (Kafka) and their consistency implications.

Fowler — Event-driven.

2Migrations & evolution

0/4▶

The hardest real-world work is changing a live system with zero downtime.

Change without breaking

Strangler fig — old and new side by side, migrate consumers incrementally.

Fowler — Strangler Fig.

Zero-downtime schema changes — expand → backfill → dual-write → switch → contract.

Stripe — Online migrations.

Dual writes & backfills; verifying parity before cutover; rollback at each step.

Stripe — Online migrations.

Data backfill at scale — throttling, idempotent jobs, progress tracking.

AWS Builders' Library.

3Tradeoff leadership

0/4▶

At staff, the answer is rarely a diagram — it's a defensible decision with cost, risk, and org impact weighed and written down.

Decide and align

Cost modeling — $ per request, storage, egress; cost as a design constraint.

Well-Architected — Cost.

Build vs buy vs adopt; reversible vs one-way-door decisions.

Werner Vogels — one-way doors.

Write a crisp design doc / RFC: context, options, tradeoffs, recommendation, risks.

Design docs at Google.

Drive alignment across teams; surface disagreement early; disagree-and-commit.

StaffEng — Communicating.

4Reliability org & DR

0/3▶

You set the reliability strategy, not just the retry policy.

Own reliability at scale

SLOs & error budgets as a cross-team planning tool; when to freeze features.

SRE Workbook — Error budgets.

Incident response & blameless postmortems; reducing MTTR.

SRE Book — Incidents.

Disaster recovery: RTO/RPO, backups, region failover drills, chaos testing.

Principles of Chaos.

5Security & compliance at scale

0/3▶

Staff designs bake in security, tenancy isolation, and compliance from the start.

Design safely

Defense in depth; least privilege; secrets management & rotation.

OWASP — Top 10.

Multi-tenant isolation, noisy neighbours, per-tenant limits.

AWS Builders' Library.

Data governance & compliance (PII, encryption at rest/in transit, audit trails).

GDPR overview.

6Ambiguous & platform prompts

0/8▶

Staff prompts are deliberately vague ("design our platform"). Frame the problem and choose what not to build.

Frame, then design

Design a multi-tenant platform (isolation, limits, billing).

AWS Builders' Library.

"Cut infra cost 40%" — measure, find the big spenders, tier storage, rightsizing.

Well-Architected — Cost.

Deprecate a core service used by 30 teams — comms, migration path, guardrails.

Strangler Fig.

Internal developer platform (CI/CD, environments, golden paths).

internaldeveloperplatform.org.

Global feature-flag / experimentation platform for many teams.

Fowler — Feature Toggles.

Company-wide data platform (ingestion → warehouse/lake → governance).

DDIA — ch. 10–12.

Unify two acquired systems with overlapping data models — sequencing & risk.

Strangler Fig.

Capacity & cost at 10× — where it breaks first and the cheapest fix.

Primer — Estimates.