Competitive Analysis for Multi-Commodity Ski-Rental Problem

Authors: Binghan Wu, Wei Bao, Dong Yuan, Bing Zhou

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical results verify our theoretical conclusion and demonstrate the advantages of MOBE in a real-world scenario.
Researcher Affiliation Academia Faculty of Engineering, The University of Sydney
Pseudocode Yes Algorithm 1: Multi-Object Break-Even (MOBE); Algorithm 2: Threshold Solver; Algorithm 3: Sub Threshold Solver
Open Source Code No The paper does not provide any explicit statement or link for the release of its source code.
Open Datasets Yes [Kamal, 2019] Sani Kamal. Forest fires in india. https: //www.kaggle.com/sanikamal/forest-fires-in-india, 2019. Accessed: 2021-09-19.
Dataset Splits No The paper describes the 'input' for its trace-driven experiment but does not specify any explicit training, validation, or testing splits for this data, as it's an online algorithm simulation rather than a typical machine learning model training scenario.
Hardware Specification No The paper mentions '8 v CPU and 16 GB memory' or '16 v CPU and 32 GB memory' but these refer to the specifications of the *cloud service components being modeled* (e.g., Fargate instances), not the actual hardware used to run the simulation or experiments.
Software Dependencies No The paper mentions 'AWS Fargate and AWS Sage Maker notebook xlarge serverless technologies' which are cloud services. However, it does not provide specific version numbers for any software libraries, frameworks, or programming languages used to implement or run the algorithm.
Experiment Setup Yes We set three typical configurations for Fargate instances: (1) web server: 8 v CPU and 16 GB memory; (2) micro service server: 16 v CPU and 32 GB memory; (3) middle ware (e.g. zookeeper, kafka) 8 v CPU and 32 GB memory. The pricing information is given in Table 2 [AWS, 2021b; AWS, 2021a]. We added up the number of forest fires from 2008 to 2011 in the data set and obtained the number of forest fires in each state in 3 years as the input. We conduct the experiment in three groups. Group 1 uses six web servers to simulate a small web application, Group 2 consists of 4 web servers, 8 micro services, and 6 middlewares to emulate a large web application, and Group 3 adds 4 Sage Maker notebooks for fire forecasting upon Group 2. In each experiment, we randomly select a state s three-year total number of forest fires from the data set.