CoRecursive: Coding Stories Tech Talk: Data and Scale with Pat Helland - The long view on distributed databases
34 snips
Mar 31, 2019 Pat Helland, a longtime expert in distributed databases and currently with Salesforce, shares his insights on the complexities of modern data storage. He explains how to navigate the world of distributed systems using relatable metaphors, like comparing messaging challenges to postal communication. Helland dives into state management in microservices, the nuances of managing distributed transactions, and the significance of immutability in data handling, all while emphasizing the balance between theoretical concepts and practical business needs.
AI Snips
Chapters
Transcript
Episode notes
Entity Transaction Boundary Example
- Pat Helland uses the example of processing a single order as an entity to explain transaction boundaries.
- Each order and its related data fit on one server to allow atomic transactions within that entity only.
Workflows with Single-Entity Transactions
- Distributed systems must design workflows as sequences of single-entity transactions connected by messaging.
- This approach tolerates scale and failure better than cross-entity distributed transactions.
Operation Intentions and Interchangeability
- In distributed systems, operations represent intentions rather than direct value changes.
- Interchangeability of resources allows operation reordering without compromising system correctness.