A practitioner's
reference.
Longer-form pieces for the decisions you keep coming back to. Five tracks, each a flow of primer, deep dive, and exercise, in the order that actually works.
The data layer.
How data lives, scales, and recovers.
Database indexing
B-tree vs hash vs LSM — where the extra disk goes, why queries fly, when to add a covering index, and when to delete one. The chapter to re-read before any schema review.
Database scaling
Replication, sharding, partitioning, and the operational reality of each — why read replicas don't fix writes, why a sharding key is one-way, and what to do when shards rebalance.
NoSQL databases
Key-value, wide-column, document, graph — the four shapes, which workloads each one fits, and the rule that decides every choice: where you want the schema enforced.
Caching, layered.
From a hashmap in front of the DB to a global edge fabric.
Caching strategies
Write-through, write-behind, cache-aside, refresh-ahead — every canonical caching strategy in one place, with the latency math and a one-line decision rule for each.
Advanced caching
The failure modes the primer left out: thundering herds, stampede protection, request coalescing, jitter, warming, and the four kinds of invalidation that bite in production.
Going horizontal.
When one box stops being enough.
Scaling out
Horizontal growth patterns and the seams they introduce — stateless tiers, session storage, sticky routing, and the moment "just add more boxes" stops being enough.
Load balancing
L4 vs L7, round-robin vs least-connections vs consistent hashing — where you choose, what you measure, and the failure modes that don't show up in marketing copy.
Monolith limits
The edge where one codebase stops being enough — the four signals that say it's time to split, and the trap of splitting before any of them have actually fired.
Distributed mechanics.
Async work, retries, and failures.
Async architecture
Events, queues, idempotence, at-least-once delivery — the four message-delivery semantics, what each costs the producer, and what each demands of the consumer.
Orchestration & resiliency
Timeouts, retries with jitter, bulkheads, circuit breakers — the four patterns every reliable distributed system implements, and the rules for when each one applies.
Network protocols & DNS
The layers, the acronyms, the practical bits — when to pick UDP, when DNS-based load balancing fails, what mTLS actually buys, and what shows up in design reviews.
Putting it on paper.
The full design exercise.
Mechanics, next door.
The handbook is about which piece to pick. The guides explain how each piece actually works. They cross-reference freely.