|
ИСТИНА |
Войти в систему Регистрация |
Интеллектуальная Система Тематического Исследования НАукометрических данных |
||
Using Empirical Data for Scalability Analysis of Parallel Applicationsстатья
Информация о цитировании статьи получена из
Scopus
Дата последнего поиска статьи во внешних источниках: 20 декабря 2019 г.
- Авторы: Valkov Pavel, Kazmina Kristina, Antonov Alexander
- Сборник: Communications in Computer and Information Science
- Том: 1063
- Год издания: 2019
- Место издания: Springer International Publishing
- Первая страница: 58
- Последняя страница: 73
- DOI: 10.1007/978-3-030-28163-2_5
- Аннотация: Scalability is the most important characteristic describing the execution of a parallel program. Scalability is a quantity that depends on many factors: the values of the input parameters, the algorithm used and its implementation details, the capabilities of hardware resources, and others. We must evaluate and compare the scalability of applications to determine their optimal execution parameters on the resources available. However, there often arises a need to assess scalability on large configurations of computing systems to which access may be limited. Therefore, an important task is to predict scalability based on the results obtained on small configurations. The concept of scalability metric was introduced for this purpose. In this article, we study the applicability of the metric and consider some approaches to predicting the scalability of applications on large supercomputer configurations. We tested the applicability of the proposed approach on the Lomonosov and Lomonosov-2 supercomputers at Lomonosov Moscow State University.
- Добавил в систему: Антонов Александр Сергеевич