System Internals Guides

How the widely used infrastructure systems actually work.

Choose a concept, then work through an example

10 guides in this section. Each article states its audience and assumptions, explains specialized terms, and gives a worked example or practical exercise.

Technical topics distinguish toy models from product behavior. Leadership topics distinguish context-dependent frameworks from evidence and personal claims.

01·Apache CassandraSeparate Cassandra's partition placement, timestamp reconciliation, consistency levels, and crash durability. 02·OpenObserveFollow telemetry from stateful ingestion to immutable analytical files and a versioned query catalog. 03·MinIOUnderstand how an object store reconstructs bytes from erasure-coded shards and safely publishes versions. 04·RedisUnderstand isolated commands, in-memory state, expiry, persistence, and asynchronous failover as separate mechanisms. 05·Apache KafkaUnderstand a partition log, replica visibility, producer retries, and consumer recovery without conflating them. 06·Apache FlinkUnderstand recoverable stream state, event-time windows, and bounded flow control. 07·ZooKeeperUnderstand versioned coordination, session liveness, ordered writes, and possibly stale reads. 08·NGINXUnderstand event-driven proxying and reloads without assuming idle connections are free or every reload is lossless. 09·ElasticsearchSeparate indexed durability, search refresh, replica state, and query completeness. 10·Apache LuceneUnderstand documents, postings, reader snapshots, and durable commits beneath search products.

Find a guide

10 guides

Apache Cassandra — Internal Architecture

Separate Cassandra's partition placement, timestamp reconciliation, consistency levels, and crash durability.

Consistent hashing · Quorum · LSM trees, memtables, and SSTables · Tombstones · Schema compatibility and semantic contracts · Linearizability

Elasticsearch — Internal Architecture

Separate indexed durability, search refresh, replica state, and query completeness.

Refresh, commit, and reader snapshots · Analyzers, postings, and doc values · Inverted indexes and columnar storage · Optimistic concurrency and compare-and-set · Offsets and checkpoints · Schema compatibility and semantic contracts

Apache Flink — Internal Architecture

Understand recoverable stream state, event-time windows, and bounded flow control.

Keyed state, checkpoints, and savepoints · Event time and watermarks · Offsets and checkpoints · Backpressure and load shedding · Cardinality

Apache Kafka — Internal Architecture

Understand a partition log, replica visibility, producer retries, and consumer recovery without conflating them.

Log offsets and commit boundaries · Producer idempotence and transactions · Schema compatibility and semantic contracts · Offsets and checkpoints · Backpressure and load shedding · Write-ahead log and fsync

Apache Lucene — Internal Architecture

Understand documents, postings, reader snapshots, and durable commits beneath search products.

Analyzers, postings, and doc values · BM25 ranking · Refresh, commit, and reader snapshots · Inverted indexes and columnar storage · Complexity and workload assumptions

MinIO — Internal Architecture

Understand how an object store reconstructs bytes from erasure-coded shards and safely publishes versions.

Reed–Solomon erasure coding · Quorum · Failure domains and redundant capacity · Multi-tenancy and publication lifetime · Content addressing and garbage collection

NGINX — Internal Architecture

Understand event-driven proxying and reloads without assuming idle connections are free or every reload is lossless.

Event loops and nonblocking I/O · Reverse proxy, buffering, and graceful reload · Cache-aside and invalidation · Backpressure and load shedding · Wall and monotonic clocks

OpenObserve — Internal Architecture

Follow telemetry from stateful ingestion to immutable analytical files and a versioned query catalog.

Multi-tenancy and publication lifetime · Write-ahead log and fsync · Partition pruning and predicate pushdown · Inverted indexes and columnar storage · Offsets and checkpoints · Rollups and weighted aggregation

Redis — Internal Architecture

Understand isolated commands, in-memory state, expiry, persistence, and asynchronous failover as separate mechanisms.

Isolation is not rollback · Snapshots, AOF, and power-loss durability · Event loops and nonblocking I/O · Wall and monotonic clocks · Cache-aside and invalidation · Quorum

Apache ZooKeeper — Internal Architecture

Understand versioned coordination, session liveness, ordered writes, and possibly stale reads.

znodes, sessions, and watches · ZAB and ZooKeeper consistency · Quorum · Fencing tokens · Leases · Optimistic concurrency and compare-and-set