Reproduce “USDB: Scaling Storage for Next-Generation Recommender Systems at ByteDance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiaoran Zhu, Jieming Cai, Weiping Yu, Xi Chen, Shu Cai, Jiahao Li, Yifan Xiao, Yan Deng and 6 more
USDB maintains user histories for recommendation serving and training by appending actions, consolidating them in the background, and sharing resulting files. Arrival-version filtering helps reconstruct the history visible at inference despite late events. In production migration, mean write latency fell from 123.94 to 17.56 microseconds, while mean read latency stayed near 410 microseconds; controlled read gains depended on request size.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.