Architecture Learning

How B-tree indexes organize data, how insertions cause node splits, how queries traverse the tree, and why indexes have a write amplification cost.

How normal forms eliminate data anomalies, when denormalization is the correct engineering tradeoff, and how to reason about read vs. write integrity at schema design time.

How PostgreSQL's query planner chooses between sequential scan, index scan, and bitmap scan; how table statistics drive cost estimates; and which index type to choose for which access pattern.

How vector embeddings enable semantic search, how HNSW and IVFFlat indexes make approximate nearest neighbor search practical at scale, and when to use pgvector versus a dedicated vector database.

How a single application write triggers multiple physical I/O operations across WAL, heap pages, and indexes, why this limits write throughput, and how LSM trees trade read amplification for lower write amplification.

How partition key design determines load distribution, how sequential keys create hotspot partitions, why hotspots cause cascading failures, and how to design partition keys that distribute load uniformly.

How primary-replica replication lag forms, accumulates under write pressure, causes stale reads, and how applications should handle it.

The consistency spectrum from linearizability to eventual consistencywhat each model guarantees, which real systems implement each model, and how to design application code for the consistency level your infrastructure provides.

How eventual consistency models propagate updates across nodes, how convergence windows create read anomalies, how conflict resolution works, and when strong consistency is required instead.

Why cache invalidation is hard, how stale data propagates, how invalidation failures cause cache poisoning, and how to design invalidation-safe patterns.

How message queues build backlogs under producer surges, how consumer lag propagates downstream, how unbounded queues become memory bombs, and how backpressure prevents cascade failures.

How Kafka consumer groups track offsets, how consumer lag accumulates when processing falls behind production, how uneven partition assignment creates lag skew, and how to instrument and remediate consumer lag before it causes data pipeline delays.

Why distributed systems cannot simultaneously provide consistency, availability, and partition tolerance: and how PACELC extends this to the latency-consistency tradeoff that applies even when the network is healthy.

How inverted indexes enable full-text search, how BM25 relevance scoring ranks results, and how Elasticsearch's indexing pipeline, refresh cycle, and shard architecture determine operational behavior under load.

How OLTP and OLAP workloads have fundamentally different access patterns, why row-oriented and column-oriented storage are each optimal for one, how mixing them on the same system causes interference, and how hybrid architectures separate the concerns.

How separating the command and query models enables independent scaling, what consistency guarantees CQRS gives up, how read model staleness propagates, and when the operational complexity is justified.

Store a sequence of immutable domain events as the source of truth; derive current state by replaying the event log. Understand projections, snapshots, event versioning, and the operational tradeoffs that make event sourcing correct in some systems and wrong in others.

How to migrate production systems incrementally using strangler fig, expand-contract schema changes, traffic shadowing, and online schema change tools: without big-bang rewrites, downtime, or data loss.

Three isolation models for multi-tenant systems: shared database with row-level security, schema-per-tenant, and database-per-tenant: with precise tradeoffs across isolation, cost, operational complexity, and scalability.

How audit logs differ from application logs, what every audit entry must contain, how to design an immutable append-only audit store, and the operational patterns for querying, partitioning, and securing audit data at scale.