| Autor: |
Gursharan Singh, Parminder Singh, Anas Motii, Mustapha Hedabou |
| Jazyk: |
angličtina |
| Rok vydání: |
2024 |
| Předmět: |
|
| Zdroj: |
Journal of King Saud University: Computer and Information Sciences, Vol 36, Iss 1, Pp 101887- (2024) |
| Druh dokumentu: |
article |
| ISSN: |
1319-1578 |
| DOI: |
10.1016/j.jksuci.2023.101887 |
| Popis: |
The present research has found that container virtualization systems are more effective and more straightforward to handle in contrast to traditional virtualization systems. The live migration processes can be substantially reduced by using lightweight containers developed on virtualization techniques. It shrinks the suspension of service by transmitting minimum memory of the source without stopping the migration process. Therefore, transferring many memory pages leads to long migration time and downtime, which affects the performance of containerized applications and cost is a major concern. In this paper, the pre-copy container live migration is analyzed in a detailed manner to trace out the possibilities to improve the performance in various applications and introduce a probability-based migration technique to overcome the memory transmission limitations. For continual learning, we used containers to run and manage the AI-based system on the cloud to overcome the delay in service during migration. The pre-dump is the first phase of sending the container file system to the destination host. Additionally, dirty pages are predicted during a migration process using the Meta-Heuristic approach. After that, the active set of pages and their update rate have also been identified. In addition, based on the threshold level of the maximum update rate, the pages have been shortlisted to be discarded from pre-dump. The proposed technique is implemented and tested on different scenarios, and results show that it reduces the size of the pre-dump by 26.48% compared to the traditional pre-copy migration technique, which leads to less downtime for cloud services. The results reveal that the suggested technique surpasses existing alternatives in overall migration time and amount of data transmitted. There is 31.04% reduction in the amount of data transferred during the iterative phase. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
