All Articles (9 articles)

  • CRDTs for Collaborative Systems

    System Design / Core Distributed Patterns 19 min read

    Conflict-free Replicated Data Types (CRDTs) are data structures mathematically guaranteed to converge to the same state across distributed replicas without coordination. They solve the fundamental challenge of distributed collaboration: allowing concurrent updates while ensuring eventual consistency without locking or consensus protocols.This article covers CRDT fundamentals, implementation variants, production deployments, and when to choose CRDTs over Operational Transformation (OT).

  • Operational Transformation

    System Design / Core Distributed Patterns 17 min read

    Deep-dive into Operational Transformation (OT): the algorithm powering Google Docs, with its design variants, correctness properties, and production trade-offs.OT enables real-time collaborative editing by transforming concurrent operations so that all clients converge to the same document state. Despite being the foundation of nearly every production collaborative editor since 1995, OT has a troubled academic history—most published algorithms were later proven incorrect. This article covers why OT is hard, which approaches actually work, and how production systems sidestep the theoretical pitfalls.

  • Distributed Locking

    System Design / Core Distributed Patterns 20 min read

    Distributed locks coordinate access to shared resources across multiple processes or nodes. Unlike single-process mutexes, they must handle network partitions, clock drift, process pauses, and partial failures—all while providing mutual exclusion guarantees that range from “best effort” to “correctness critical.”This article covers lock implementations (Redis, ZooKeeper, etcd, Chubby), the Redlock controversy, fencing tokens, lease-based expiration, and when to avoid locks entirely.

  • Exactly-Once Delivery

    System Design / Core Distributed Patterns 23 min read

    True exactly-once delivery is impossible in distributed systems—the Two Generals Problem (1975) and FLP impossibility theorem (1985) prove this mathematically. What we call “exactly-once” is actually “effectively exactly-once”: at-least-once delivery combined with idempotency and deduplication mechanisms that ensure each message’s effect occurs exactly once, even when the message itself is delivered multiple times.

  • Event Sourcing

    System Design / Core Distributed Patterns 21 min read

    A deep-dive into event sourcing: understanding the core pattern, implementation variants, snapshot strategies, schema evolution, and production trade-offs across different architectures.

  • Change Data Capture

    System Design / Core Distributed Patterns 20 min read

    Change Data Capture (CDC) extracts and streams database changes to downstream systems in real-time. Rather than polling databases or maintaining dual-write logic, CDC reads directly from the database’s internal change mechanisms—transaction logs, replication streams, or triggers—providing a reliable, non-invasive way to propagate data changes across systems.This article covers CDC approaches, log-based implementation internals, production patterns, and when each variant makes sense.

  • Database Migrations at Scale

    System Design / Core Distributed Patterns 15 min read

    Changing database schemas in production systems without downtime requires coordinating schema changes, data transformations, and application code across distributed systems. The core challenge: the schema change itself takes milliseconds, but MySQL’s ALTER TABLE on a 500GB table with row locking would take days and block all writes. This article covers the design paths, tool mechanisms, and production patterns that enable zero-downtime migrations.

  • Multi-Region Architecture

    System Design / Core Distributed Patterns 19 min read

    Building systems that span multiple geographic regions to achieve lower latency, higher availability, and regulatory compliance. This article covers the design paths—active-passive, active-active, and cell-based architectures—along with production implementations from Netflix, Slack, and Uber, data replication strategies, conflict resolution approaches, and the operational complexity trade-offs that determine which pattern fits your constraints.

  • Graceful Degradation

    System Design / Core Distributed Patterns 20 min read

    Graceful degradation is the discipline of designing distributed systems that maintain partial functionality when components fail, rather than collapsing entirely. The core insight: a system serving degraded responses to all users is preferable to one returning errors to most users. This article covers the pattern variants, implementation trade-offs, and production strategies that separate resilient systems from fragile ones.