Predicting the Future of Distributed Systems
summary
Object storage moves from archival to primary storage for transactional systems with concrete examples: S3: 19 years old, 4 nines availability, 11 nines durability, architectures like Aurora, Snowflake, WarpStream, SlateDB, InfluxDB disaggregate compute from storage using Parquet and DuckDB-style query libraries, and new programming models compete — Akka, Temporal, Wasm Cloud, Golem/Unison — while Bezos-style one-way/two-way door thinking guides choices and AI operationalization becomes the risk-tolerant adoption vector.
AI-written summary · about
chapters (13)
- 0:00Predicting distributed systems future
- 1:54Why predicting matters
- 3:39One-way vs two-way door decisions
- 5:04Object storage revolution
- 11:41Databases disaggregation
- 14:10Standardization in data formats
- 17:20Edge and cloud convergence
- 19:05Programming models challenges
- 22:15Abstraction layers portability
- 23:59Durable execution frameworks
- 26:27AI ops as distributed engineering
- 29:16Risks in new models
- 30:41Summary of predictions
related talks
35mBuilding Systems that Last
31mFuture Data Platforms and PaaS Services
47mA Distributed, Scalable and Highly Reliable Data Storage System
39mHigh-volume Data Streaming and Storage
42mPostgreSQL Distributed: Architectures & Best practices
29mEnhancing Flexibility and Productivity with Access Patterns and Storage-Agnostic Abstractions
45mDo Distributed Systems they said. It'll be fun they said.
31mThe Design of a Modern Distributed Object Storage