SkipPredict: When to Invest in Predictions for Scheduling

Authors: Rana Shahout, Michael Mitzenmacher

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To gain more insight into when to invest in prediction and how Skip Predict compares to other policies, in this section we compare Skip Predict, SPRPT, 1bit and FCFS using simulation with realworld and synthetic traces in the setting of the single queue with Poisson arrivals.
Researcher Affiliation Academia Rana Shahout Harvard University Michael Mitzenmacher Harvard University
Pseudocode No The paper describes algorithms verbally and mathematically but does not include any explicitly labeled "Pseudocode" or "Algorithm" blocks with structured code-like steps.
Open Source Code Yes We provide our code and scripts to reproduce all the results in the paper and the Appendix with readme file contains instructions as supplemental material.
Open Datasets Yes For real-world traces, we used three traces from Amvrosiadis et al. (Amvrosiadis et al., 2018): Twosigma, Google, and Trinity.
Dataset Splits No The paper describes the overall simulation setup and how mean response times are calculated over many runs, but it does not specify explicit training, validation, and test dataset splits with percentages or sample counts in the typical machine learning sense.
Hardware Specification Yes The evaluation was performed on an AMD EPYC 7313 16-Core Processor running Ubuntu 20.04.6 LTS with Linux kernel 5.4.0 172-generic.
Software Dependencies Yes We implemented the simulation in Python 3.7.6.
Experiment Setup Yes The default costs for the external model are c1 = 0.5, c2 = 20, and in the server time are c1 = 0.05, c2 = 0.5.