Prediksi Query Traces Dari RocksDB Menggunakan Machine Learning

Februanto, Theodorus Wensan (2024) Prediksi Query Traces Dari RocksDB Menggunakan Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07211940000025-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 July 2026. Download (10MB) | Request a copy

Abstract

RocksDB, salah satu basis data key-value, memainkan peran penting dalam berbagai sistem industri dan akademis. Salah satu fitur penting dari RocksDB ialah fitur tracing. Fitur tracing ini dapat menghasilkan beberapa berkas berupa query traces. Query traces ini berisi catatan aktivitas dan aliran eksekusi dari basis data RocksDB. Catatan ini memiliki banyak kegunaan dan manfaat, terutama untuk keperluan industri dan riset. Sayangnya, tracing pada RocksDB sulit untuk dilakukan karena adanya performance overhead pada saat melakukan tracing. Untuk menghilangkan performance overhead ini, penulis mengusulkan ide untuk membuat query traces sintetis dari RocksDB dengan menggunakan machine learning (ML). Untuk membuat query traces sintetis ini, ML akan memprediksi query traces dari RocksDB. Akurasi dan R-squared yang didapat dari percobaan ini ialah 82.14% dan 74.60%, membuktikan bahwa query traces sintetis yang dihasilkan cukup mirip dengan query traces organik dari RocksDB, serta dapat digunakan dengan semestinya.
===================================================================================================================================
RocksDB, one of the key-value databases, plays an important role in various industrial and academic systems. One of the important features of RocksDB is the tracing feature. This tracing feature can generate several files, which are query traces. These query traces contain records of activities and execution flows of a RocksDB database. These records have many uses and benefits, especially for industrial and research purposes. Unfortunately, tracing on RocksDB is difficult to perform because of the performance overhead that happens while performing tracing. To eliminate this performance overhead, we propose a novel idea which is creating synthetic query traces from RocksDB using machine learning. To create these synthetic query traces, machine learning will predict the query traces of RocksDB. The accuracy and R-squared obtained from this experiment are 82.14% and 74.60%, proving that the synthetic query traces are quite similar to the organic query traces and can be used properly.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	RocksDB, Basis Data, Key-Value, Tracing, Traces, Query Traces, Performance Overhead, Machine Learning, Prediksi; RocksDB, Database, Key-Value, Tracing, Traces, Query Traces, Performance Overhead, Machine Learning, Prediction.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.758 Software engineering T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Theodorus Wensan Februanto
Date Deposited:	15 Feb 2024 03:26
Last Modified:	15 Feb 2024 03:26
URI:	http://repository.its.ac.id/id/eprint/107049

Actions (login required)

View Item