Reproduce “SPLIT: SymPathy for Large jobs Improves Tail latency” 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

Zhouzi Li, Mor Harchol-Balter, Alan Scheller-Wolf

The paper introduces SPLIT, a scheduler for heavy-tailed jobs across multiple servers. It assigns one server to the largest job while the others serve jobs with the shortest remaining times. In simulations at load ρ=0.5, SPLIT beats SRPT-n at the 99.9th and 99.99th percentiles, but not the 99th. Its proved strong tail optimality applies only when ρ < (n−1)/n.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.