발간년도 : [2022]
| 논문정보 |
|
| 논문명(한글) |
[Vol.17, No.5] Design and Implementation of Metadata De-Duplication Controller Systems Based on Server Storage |
|
| 논문투고자 |
Yeong-A Kim, Hyun-Ju Kim, Chang-Geun Kim |
|
| 논문내용 |
Server redundancy is typically seen in IT environments that rely on large-scale server reliability. For servers, parallelization that applications can use to justify the amount of parallel hardware should be considered. Independent computers, standard local area networks, and computer clusters formed by switches may be considered. With the advent of the Fourth Industrial Revolution, digital data has reached saturation due to the spread of the Internet of Things. As the amount of data increases, public cloud platforms that manage structured and unstructured data are growing on platforms that expand based on servers. Platforms based on object storage incur I/O overhead due to metadata extensions based on individual objects rather than directory tree structures. In this paper, we propose a metadata deduplication controller system between servers that are deleted according to the metadata update frequency. The proposed system converts duplicate files of user files into fixed-length data chunks by extending the metadata of Linux system files.It realizes improved capacity efficiency that maintains performance by deleting only chunks with low update frequency. Additionally, the system's file metadata is flexible for scaling and replication between servers. The proposed system consists of the process of processing and accessing user data in the kernel driver and the process of managing deduplication metadata. Experiments have shown that the proposed method has 12.3% higher performance and deduplication rates than traditional object storage-based systems. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
