|
ИСТИНА |
Войти в систему Регистрация |
Интеллектуальная Система Тематического Исследования НАукометрических данных |
||
Application of recommender systems approaches to the MPI program execution time predictionстатья
Информация о цитировании статьи получена из
Scopus
Дата последнего поиска статьи во внешних источниках: 15 сентября 2021 г.
- Авторы: Chupakhin Andrey, Kolosov Alexey, Bahmurov Anatoly, Antonenko Vitaly, Ishelev Gleb
- Сборник: 3 International Scientific and Technical Conference Modern Computer Network Technologies (MoNeTeC-2020)
- Серия: IEEE Conference Publications
- Год издания: 2020
- Издательство: Institute of Electrical and Electronics Engineers
- Местоположение издательства: Piscataway, NJ, United States
- DOI: 10.1109/MoNeTeC49726.2020.9258345
- Аннотация: The problem of MPI programs execution time predictionon a certain set of computer installations is considered.This problem emerge with orchestration and provisioning avirtual infrastructure in a cloud computing environment over aheterogeneous network of computer installations: supercomputersor clusters of servers (e.g. mini data centers). One of the keycriteria for the effectiveness of the cloud computing environmentis the time staying by the program inside the environment.This time consists of the waiting time in the queue and theexecution time on the selected physical computer installation, towhich the computational resource of the virtual infrastructure isdynamically mapped. One of the components of this problem isthe estimation of the MPI programs execution time on a certainset of computer installations. This is necessary to determine aproper choice of order and place for program execution. Thearticle proposes two new approaches to the program executiontime prediction problem. The first one is based on computerinstallations grouping based on the Pearson correlation coefficient.The second one is based on vector representations ofcomputer installations and MPI programs, so-called embeddings.The embedding technique is actively used in recommendationsystems, such as for goods (Amazon), for articles (Arxiv.org), forvideos (YouTube, Netflix). The article shows how the embeddingstechnique helps to predict the execution time of a MPI programon a certain set of computer installations.
- Добавил в систему: Антоненко Виталий Александрович