Peningkatan Kinerja Kueri Data Warehouse Menggunakan Mekanisme Partisi Memory-Aware Berbasis PROADAPT

Mulia, Hafidz (2026) Peningkatan Kinerja Kueri Data Warehouse Menggunakan Mekanisme Partisi Memory-Aware Berbasis PROADAPT. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221022-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Pertumbuhan data historis meningkatkan kebutuhan analitik jangka panjang pada data warehouse. Kinerja kueri dipengaruhi oleh konfigurasi chunk time interval dan kebijakan kompresi yang menentukan susunan data historis pada TimescaleDB. Penelitian ini mengadaptasi PROADAPT pada PostgreSQL/TimescaleDB single-VM. Tahap offline menilai kandidat konfigurasi menggunakan fungsi utilitas yang menggabungkan latensi, efektivitas chunk pruning, memory score berbasis BMR dan WSS, storage size, serta penalti overhead. Tahap online menjalankan lima window dengan persentase predefined query sebesar 100%, 75%, 50%, 25%, dan 0%. Hasil tahap offline memilih interval 3 hari pada Skenario 1 dengan utilitas 0,4056, kombinasi interval 14 hari dan compress after 7 hari pada Skenario 2 dengan utilitas 0,4700, serta kombinasi interval 7 hari dan compress after 3 hari pada Skenario 3 dengan utilitas 0,3065. Baseline menghasilkan latensi rata-rata 503,562 ms. Tahap online pada Skenario 1, Skenario 2, dan Skenario 3 menghasilkan latensi rata-rata 522,526 ms, 397,256 ms, dan 377,641 ms. BMR ketiga skenario mendapat hasil yang lebih rendah daripada baseline.
==================================================================================================================================
The growth of historical data increases the demand for long-term analytics in data warehouses. Query performance is affected by the chunk time interval and compression policy that determine how historical data are organized in TimescaleDB. This study adapts PROADAPT to a PostgreSQL/TimescaleDB single-VM environment. The offline stage evaluates configuration candidates using a utility function that combines latency, chunk-pruning effectiveness, a BMR- and WSS-based memory score, storage size, and an overhead penalty. The online stage executes five windows containing 100%, 75%, 50%, 25%, and 0% predefined queries. The offline stage selects a 3-day interval in Scenario 1 with a utility of 0.4056, a 14-day interval with a 7-day compress after in Scenario 2 with a utility of 0.4700, and a 7-day interval with a 3-day compress after in Scenario 3 with a utility of 0.3065. The baseline produces an average latency of 503.562 ms. The online stages over Scenarios 1, 2, and 3 produce average latencies of 522.526 ms, 397.256 ms, and 377.641 ms. All three scenarios produce lower BMR and higher WSS than the baseline. These results show that the configuration changes have different effects on latency and block usage as the query composition changes.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Data Warehouse, TimescaleDB, PROADAPT, Partisi Adaptif, Kompresi Memory-Aware, Data Warehouse, TimescaleDB, PROADAPT, Adaptive Partitioning, Memory-Aware Compression
Subjects:	Q Science > QA Mathematics > QA76.9.D33 Data compression (Computer science) Q Science > QA Mathematics > QA76.9.D37 Data warehousing.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Hafidz Mulia
Date Deposited:	31 Jul 2026 02:27
Last Modified:	31 Jul 2026 02:27
URI:	http://repository.its.ac.id/id/eprint/139317

Actions (login required)

View Item