Paper
1 February 2024 Analyzing performance degradation for wide stripe erasure codes
Xiangrui Yang, Jie Yao, Qiang Cao
Author Affiliations +
Proceedings Volume 13067, International Workshop on Holography and Related Technologies (IWH2022 & 2023); 130670E (2024) https://doi.org/10.1117/12.3024709
Event: International Workshop on Holography and Related Technologies (IWH2022&2023), 2023, Fuzhou, China
Abstract
In the digital age, the volume of data is expanding exponentially, rendering it as a fundamental asset. Consequently, the need for storage systems that possess substantial capacity, affordability, exceptional performance, and unwavering dependability has become increasingly pressing. In response to this demand, storage systems have extensively embraced erasure codes, particularly the wide stripe erasure codes, owing to their remarkable storage efficiency and reliability. However, this paper found that the encoding and decoding performance of wide stripe erasure codes significantly deteriorates compared to narrow stripe erasure codes, when using mainstream erasure code acceleration libraries. In order to gain a comprehensive understanding of this phenomenon, extensive experimentation was conducted to test the performance of encoding and decoding process of erasure codes. Further, hardware events during erasure code calculation were analyzed. The results show that the root cause of performance degradation is the increasing L3 cache misses which saw a 240% increase given a set amount of encoding data
(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.
Xiangrui Yang, Jie Yao, and Qiang Cao "Analyzing performance degradation for wide stripe erasure codes", Proc. SPIE 13067, International Workshop on Holography and Related Technologies (IWH2022 & 2023), 130670E (1 February 2024); https://doi.org/10.1117/12.3024709
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Reliability

Data storage

Statistical analysis

Analytical research

Clouds

Computer hardware

Forward error correction

Back to Top