Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM64460
Volume:
10 |
6 |
Jun 2026
Pages:
1-4
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
ASET
Amity University Chhattisgarh
Department of Computer Science and Engineering
Amity University Chhattisgarh Raipur, India
Cache eviction policies govern which entries a bounded in-memory store discards when capacity is exhausted. Prevailing production systems rely on deterministic heuristics—Least Recently Used (LRU) and Least Frequently Used (LFU)—each of which interrogates a single behavioural dimension and therefore mischaracterises workloads where temporal recency and cumulative frequency carry different predictive weight across key populations. This paper introduces PyVault, a lightweight key-value store constructed entirely in Python, whose eviction subsystem replaces static heuristics with a supervised Random Forest classifier trained on a five-dimensional feature vector derived from a continuously maintained per-key access log. The classifier assigns each resident key a probabilistic coldness score; when eviction is triggered, the store removes the highest-scoring keys rather than the oldest or least-counted ones. Controlled experiments across five synthetic workloads modelled on Zipf-distributed access sequences show that the proposed approach raises cache hit rates by 3.32 to 18.32 percentage points relative to LRU and by 2.44 to 16.66 percentage points relative to LFU, with the largest absolute gain observed under a balanced-capacity regime. The system additionally provides disk persistence via atomic snapshot writes, a RESTful HTTP interface with header-based authentication, an auto-retraining loop that refreshes the classifier every sixty seconds, and a browser-based monitoring dashboard. The complete artefact is packaged as a Docker container for reproducible cross-platform deployment.
KEYWORDS
We have presented PyVault, an in-memory key-value store that supplants static eviction heuristics with a supervised Random Forest classifier trained on a continuously updated per-key access log. Controlled experiments across five synthetic workloads demonstrate that the proposed approach yields cache hit-rate improvements of 3.32 to 18.32 percentage points over LRU and 2.44 to 16.66 percentage points over LFU, with the largest gains observed under workload regimes where the hot-key cohort is small relative to total capacity and traffic skew is moderate. The system is complete with atomic persistence, RESTful access, API-key authentication, a browser-based monitoring dashboard, and Docker packaging for reproducible deployment. The source code and benchmark scripts are available in the project repository.
[1] W. A. King, "Analysis of paging algorithms," in Proc. IFIP Congress, 1971, pp. 485–490. [2] R. L. Mattson, J. Gecsei, D. R. Slutz, and I. L. Traiger, "Evaluation techniques for storage hierarchies," IBM Syst. J., vol. 9, no. 2, pp. 78–117, 1970. [3] M. Arlitt and T. Jin, "A workload characterization study of the 1998 World Cup web site," IEEE Netw., vol. 14, no. 3, pp. 30–37, May–Jun. 2000. [4] N. Megiddo and D. S. Modha, "ARC: A self-tuning, low overhead replacement cache," in Proc. 2nd USENIX Conf. File Storage Technol. (FAST), 2003, pp. 115–130. [5] S. Jiang and X. Zhang, "LIRS: An efficient low inter-reference recency set replacement policy to improve buffer cache performance," ACM SIGMETRICS Perform. Eval. Rev., vol. 30, no. 1, pp. 31–42, Jun. 2002. [6] J. Shi, C. Cher, J. Torrellas, and X. Shen, "Reuse distance analysis for performance prediction of irregular programs," ACM Trans. Archit. Code Optim., vol. 9, no. 4, pp. 1–27, Jan. 2013. [7] G. Vietri, N. Balmau, K. Argyraki, and W. Zwaenepoel, "Driving cache replacement with ML-based LeCaR," in Proc. 10th USENIX Workshop Hot Topics Storage File Syst. (HotStorage), 2018. [8] H. Shi, C. Zhang, J. Yang, S. Zhou, Y. Zhu, and Y. Xiao, "Learned caching: Using machine learning for cache eviction prediction in large-scale CDN," in Proc. IEEE INFOCOM, 2022, pp. 1–10. [9] E. Lykouris and S. Vassilvitskii, "Competitive caching with machine learned advice," in Proc. 35th Int. Conf. Mach. Learn. (ICML), vol. 80, 2018, pp. 3296–3305. [10] Redis Ltd., "Redis documentation: Memory management and eviction policies," 2024. [Online]. Available: https://redis.io/docs/reference/eviction/
Publication Date
6/3/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
4
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...