// papershelf — 3 papers

Papers worth the second read.

Mostly systems and retrieval. Each entry has the claim, the mechanism, and where it shows up in something I've shipped.

USENIX ATC 2014 · 2014

In Search of an Understandable Consensus Algorithm (Raft)

Diego Ongaro, John Ousterhout

Leader election, log replication and safety, decomposed so a human can hold each in their head. Understandability is a design goal, not a side effect.

#distributed-systems#consensus#databases

SOSP 2007 · 2007

Dynamo: Amazon's Highly Available Key-value Store

Giuseppe DeCandia, Deniz Hastorun, Madan Jampani et al.

Availability over consistency is a business decision made explicit: a shopping cart that always accepts writes beats one that is occasionally correct and down.

#distributed-systems#databases#dynamodb#availability

NeurIPS 2020 · 2020

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

Patrick Lewis, Ethan Perez, Aleksandra Piktus et al.

Parametric memory (the model) plus non-parametric memory (an index) beats either alone on knowledge-heavy tasks, and the index can be updated without retraining.

#rag#ai#retrieval#llms