Szczegóły publikacji
Opis bibliograficzny
Long-term prediction of cloud resource usage in high-performance computing / Piotr NAWROCKI, Mateusz Smendowski // W: Computational Science – ICCS 2023 : 23rd international conference : Prague, Czech Republic, July 3–5, 2023 : proceedings, Pt. 2 / eds. Jiří Mikyška [et al.]. — Cham, Switzerland : Springer, cop. 2023. — (Lecture Notes in Computer Science ; ISSN 0302-9743 ; LNCS 14074). — ISBN: 978-3-031-36020-6; e-ISBN: 978-3-031-36021-3. — S. 532–546. — Bibliogr., Abstr. — Publikacja dostępna online od: 2023-06-26
Autorzy (2)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 147728 |
|---|---|
| Data dodania do BaDAP | 2023-07-20 |
| DOI | 10.1007/978-3-031-36021-3_53 |
| Rok publikacji | 2023 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Springer |
| Konferencja | 23rd International Conference on Computational Science |
| Czasopismo/seria | Lecture Notes in Computer Science |
Abstract
Cloud computing is gaining popularity in the context of high-performance computing applications. Among other things, the use of cloud resources allows advanced simulations to be carried out in circumstances where local computing resources are limited. At the same time, the use of cloud computing may increase costs. This article presents an original approach which uses anomaly detection and machine learning for predicting cloud resource usage in the long term, making it possible to optimize resource usage (through an appropriate resource reservation plan) and reduce its cost. The solution developed uses the XGBoost model for long-term prediction of cloud resource consumption, which is especially important when these resources are used for advanced long-term simulations. Experiments conducted using real-life data from a production system demonstrate that the use of the XGBoost model developed for prediction allowed the quality of predictions to be improved (by 16%) compared to statistical methods. Moreover, techniques using the XGBoost model were able to predict chaotic changes in resource consumption as opposed to statistical methods.