Distributed Systems

Introductory chapter: why we need consistency models, consensus and working with failures.

Fundamental limitation of distributed systems: consistency, availability, resistance to partitioning. History, misconceptions, ACID vs BASE.

Extending CAP: Tradeoffs between latency and consistency in native mode. System classification: PA/EL, PC/EC, PA/EC, PC/EL.

Practice of time synchronization: physical vs logical clocks, NTP/PTP, clock skew impact and architectural protection against time drift.

How to design leader election: leases, quorum, failover, split-brain protection and practical implementations on Raft/ZooKeeper/etcd/Kubernetes.

How systems negotiate a single meaning: quorums, two-phase Paxos, and leader-centric Raft.

How Lamport's ideas (happens-before, logical clocks, Paxos, TLA+) grew out of physics and why they are critical for modern distributed systems.

Practical analysis of distributed transactions: coordinator, prepare/commit phases, failure modes, blocking trade-offs and alternatives via Saga/outbox.

Distributed systems testing project: hierarchy of consistency models, Serializable vs Linearizable, known findings.

A practical approach to testing distributed systems: chaos engineering, contract testing and integration testing at scale.

Analysis of the book by Martin Kleppmann: data models, replication, partitioning, transactions, batch and stream processing.

The seminal work of Tanenbaum and van Steen: architectures, coordination, consistency, fault tolerance and security.

Evolution of the Google network from the internet/streaming/cloud era to the AI-era: WAN as new LAN, multi-shard design, Protective ReRoute, intent-driven programmability and autonomous operations.

Andrew Psaltis about stream processing: Collection/Queue/Analysis tiers, delivery semantics, data windows, stream algorithms.

Distributed stream processing platform: producers, consumers, partitions, replication, delivery semantics and Kafka Streams.

A single flow circuit without a separate batch layer: immutable log, materialized views, replay/backfill and comparison with Lambda.

How to design a data pipeline: batch + streaming, ETL vs ELT, orchestration, data quality, recovery, cost control and operational reliability.

Practical analysis of Apache Iceberg: snapshots, manifests, ACID in the data lake, schema evolution, hidden partitioning, time travel and the place of Tableflow in the streaming circuit.

Nathan Marz about Lambda Architecture: batch/serving/speed layers, data immutability, HyperLogLog and practical examples.

A practical guide to adopting data mesh: domain ownership, data as a product, federated computational governance, self-serve platforms, and an MVP in one month.

Evolution of T-Bank's data platform: from DWH approaches to Lakehouse, key contours of the platform, scale and practical architectural conclusions.

Research Insights Made Simple #6: centralization vs federalization, data mesh in practice, OLTP/MPP limitations and the evolution of data platforms.

A short documentary about the local-first approach: offline experience, synchronization and user control of data.