Murat Buffalo's blog provides insights into computer science research, machine learning, and artificial intelligence. Readers can explore articles covering topics such as algorithm design, data mining, and computational biology. Additionally, they can learn about machine learning algorithms, deep learning techniques, and applications of AI in various domains.

Metadata

A review of a 2008 research paper that proposed building a relational database on Amazon S3, using SQS as a write-ahead log and S3 as a page store. The design pioneered the storage-compute separation philosophy now central to cloud-native databases like Aurora, Snowflake, and Delta Lake. Key challenges included SQS's non-FIFO delivery requiring idempotent log records, an atomicity protocol for all-or-nothing commits, B-link trees for lock-free reads on stale S3 pages, and weak isolation guarantees where last-writer-wins replaces traditional ANSI SQL isolation. The paper's claim that snapshot isolation requires a centralized counter is noted as outdated given modern hybrid logical clocks. Despite its clunky 2008-era protocols, the paper is credited as a conceptual precursor to modern data lake and lakehouse architectures.

Building a Database on S3

Surviving SQS and building B-link Trees on S3