The BugBash Podcast Re-Designing Data-Intensive Applications: The Shift to Cloud-Native Storage
29 snips
Feb 18, 2026 
Chris Riccomini, an engineer who builds distributed systems and startups, and Martin Kleppmann, a researcher and author on distributed systems, discuss rebuilding databases on cloud-native object storage. They explore how S3-like stores change system assumptions. They revisit CAP and propose offline availability. They debate using formal methods, model checking, and LLMs as test oracles for migrations.
AI Snips
Chapters
Books
Transcript
Episode notes
Outsource Ops To Reduce Pager Duty
- Offload operational burden by using cloud services to avoid being on pager duty for low-level failures.
- Accept trade-offs: outsourced ops reduce some toil but introduce new operational debugging needs.
Atomic Primitives Simplify Cloud Databases
- Strong consistency or compare-and-swap primitives in object stores simplify higher-level system design.
- These primitives let developers outsource consensus and build primitives (fencing, queues, WAL) directly on object storage.
Building A KV Store On Object Storage
- Chris built a key-value store on object storage inspired by TurboPuffer and WarpStream.
- He discovered many primitives (compaction queue, transactional fencing, WAL) become trivial once object stores provide CAS.
