Penerapan Model Machine Learning Untuk Meningkatkan Efisiensi Algoritma Clock Pada Cache Replacement

Al-Araby, Muhammad Haekal Muhyidin (2026) Penerapan Model Machine Learning Untuk Meningkatkan Efisiensi Algoritma Clock Pada Cache Replacement. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024221030-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (752kB)

Abstract

Penelitian ini mengusulkan pengembangan algoritma penggantian cache CLOCK yang diperkaya dengan pendekatan machine learning ntuk meningkatkan efisiensi keputusan eviction. Berbeda dari CLOCK konvensional yang bergantung pada bit referensi dan promosi halaman secara heuristik, studi ini memanfaatkan model supervised learning untuk memprediksi kelayakan objek dipertahankan di cache berdasarkan karakteristik akses yang diamati. Evaluasi dilakukan melalui simulasi trace replay menggunakan perangkat lunak kustom berbasis LibCacheSim, dengan pengukuran kinerja mencakup hit ratio, miss ratio, serta jumlah promosi. Hasil menunjukkan bahwa integrasi model machine learning mampu menekan jumlah promosi cache tanpa menurunkan kinerja hit/miss secara signifikan, sehingga perilaku penggantian menjadi lebih selektif dan efisien dibandingkan algoritma CLOCK standar. Pendekatan ini memberikan arah praktis untuk meningkatkan kebijakan cache eviction dengan memanfaatkan prediksi berbasis data pada sistem caching modern.
================================================================================================================================
This research proposes a CLOCK cache replacement algorithm enhanced with machine learning to improve the efficiency of eviction decisions. Unlike conventional CLOCK, which relies on reference bits and heuristic promotion behavior, the proposed approach uses a supervised learning model to predict whether an object should be retained in the cache based on observed access characteristics. The method is evaluated via trace replay simulation using custom software built on LibCacheSim. Performance is assessed using hit ratio, miss ratio, and the number of promotions. Results show that integrating machine learning can substantially reduce cache promotions without significantly degrading hit/miss performance, making eviction behavior more selective and efficient than standard CLOCK. These findings demonstrate the practical potential of data-driven prediction to improve cache eviction policies in modern caching systems.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Cache, Cache Eviction, CLOCK, Machine Learning, Supervised Learning, Trace Replay, LibCacheSim
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.64 Information resources management
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Muhammad Haekal Muhyidin Al-araby
Date Deposited:	20 Jan 2026 01:28
Last Modified:	20 Jan 2026 01:28
URI:	http://repository.its.ac.id/id/eprint/129755

Actions (login required)

View Item