Transactions and Consistency Trade-offs

#131 - Data Essentials in Software Architecture - Pramod Sadalage

Topic 6 of 8

Published May 1, 2023 · Clip 37:32–47:03 (9:31)

Pramod argues transaction and consistency requirements are often tech-imposed and should be business-driven: ask how much the business will invest and what delay it tolerates. He covers per-write consistency levels in distributed databases (quorum, one node, MongoDB-like options) and advises encoding conventions at tech-lead/architecture level so developers don't decide per write. For writing to two different databases (e.g., relational plus graph), he discusses eventual consistency versus transaction coordinators and two-phase commit, framing recommendations as explicit trade-offs with the business. He then responds to the host's point about ACID being simpler for developers, explaining that ACID trades off scaling and that trade-offs should be made visible, ideally through pairing to expose biases.