@msimoni: One thing I keep thinking about: with S3-like object storage as primitive, you could build a transactional database wit…
Summary
A tweet discusses the idea of building a transactional database with infinite throughput using S3-like object storage and content-addressing, where blocks are written in parallel and the root hash is updated periodically.
Similar Articles
@m_sirovatka: KV Cache re-use is the most important thing for agentic rollouts. We've integrated Mooncake Store into prime-rl with vL…
vLLM integrates Mooncake Store for distributed KV cache reuse, enabling cross-node prefix caching to efficiently serve agentic workloads with high token reuse.
@PlanetScale: Sick of slow databases? Our locally-attached NVMe storage unlocks unlimited IOPS, bringing data center performance to t…
PlanetScale announces high-performance cloud hosting for Vitess and Postgres using locally-attached NVMe storage to deliver unlimited IOPS and horizontal scaling.
@MrCollison: The craziest thing about building a game engine is I've learned a concept that scared me the most: Columnar Stores It's…
The author shares a personal breakthrough in understanding columnar stores while building a game engine, relating it to Trizen's ECS and ClickHouseDB.
SQLite is all you need for durable workflows
This blog post argues that SQLite, combined with Litestream for async backups, provides a simple and effective approach to durable execution for many workflow systems, especially AI agents, without needing a separate orchestration tier or network database.
How to Write to SSDs
This paper proposes out-of-place write optimizations for database systems to fully leverage SSD performance, achieving 1.65-2.24x throughput improvement and 6.2-9.8x reduction in flash writes on OLTP benchmarks.