Analisis Data Penggunaan Block Storage Untuk Rekomendasi Penyeimbangan Beban Kerja Aplikasi Telekomunikasi Menggunakan Klasterisasi

Sembiring, Fred Erick (2025) Analisis Data Penggunaan Block Storage Untuk Rekomendasi Penyeimbangan Beban Kerja Aplikasi Telekomunikasi Menggunakan Klasterisasi. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032211073-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Perkembangan industri teknologi informasi menghadirkan solusi infrastruktur penyimpanan data seperti block storage. PT XYZ, perusahaan telekomunikasi, menggunakan 32 server block storage untuk memproses data aplikasi. Pertumbuhan data dan variasi aplikasi menyebabkan server block storage beroperasi dengan kemampuan pemrosesan yang beragam. Server block storage ini krusial untuk meningkatkan mutu layanan aplikasi. Meskipun beban kerja block storage dapat diukur, PT XYZ belum memiliki alat untuk menganalisis data performanya. Oleh karena itu, diperlukan metoda analisis penggunaan server block storage untuk mengetahui beban kerja aplikasi telekomunikasi. Penelitian ini mengusulkan pendekatan klasterisasi untuk menganalisis server block storage PT XYZ berdasarkan profil kinerja, bertujuan mengkategorikannya ketingkatan beban kerja rendah, menengah, dan tinggi. Secara spesifik, dilakukan analisis komparatif antara metode K-Means dan DBSCAN. Lalu dilakukan evaluasi kinerja klasterisasi menggunakan metoda ground truth dengan membandingkan keluarannya terhadap penelitian sebelumnya. Dataset diambil dari data beban kerja 32 server block storage yang mencakup metrik IOPS (IOs/Sec), Service Time (ms), dan Bandwidth (KBs/Sec) selama periode 1.5 tahun. Metode Klasterisasi K-Means dan DBSCAN diterapkan pada dataset ini untuk membentuk klaster server. K-Means dengan algoritma partisional dan mengelompokkan server ke dalam K klaster (misal, K=3 untuk tiga tingkatan kinerja) berdasarkan kedekatan karakteristik. DBSCAN dengan algoritma berbasis kepadatan untuk mengidentifikasi klaster berbentuk fleksibel dan mendeteksi outlier. Kualitas hasil klasterisasi divalidasi secara eksternal dengan membandingkannya terhadap ground truth. Performa setiap algoritma diukur melalui tingkat akurasi, yaitu dengan memeriksa kesesuaian label klaster pada setiap ID server. Lalu dilakukan perbandingan klaster mana yang paling akurat. Hasil Perbandingan akurasi klaster menunjukkan DBSCAN yang paling akurat. Hasil analisis klusterisasi DBSCAN mampu digunakan untuk analisis kinerja server block storage dengan klasterisasi beban kerja yang terbagi menjadi 3 yaitu klaster beban kerja rendah, klaster beban kerja menengah dan klaster beban kerja tinggi. Kasterisasi ini diharapkan dapat memberikan rekomendasi terhadap penyeimbangan beban kerja guna mendukung optimalisasi penggunaan sumber daya TI secara lebih efisien. =====================================================================================================================================
The development of industrial information technology presents data storage infrastructure solutions such as block storage. PT XYZ, a telecommunications company, uses 32 block storage servers to process application data. Data growth and application variations cause block storage servers to operate with diverse processing capabilities. These block storage servers are crucial for improving application service quality. Although block storage performance can be measured, PT XYZ does not yet have tools to analyze its performance data. Therefore, it is necessary to analyze the block storage server workload for application load shifting recommendations. This research proposes a clustering approach to analyze PT XYZ's block storage servers based on performance profiles, with the aim of categorizing them into low, medium, and high workload levels. Specifically, a comparative analysis between the K-Means and DBSCAN methods was conducted. Then, the clustering performance was evaluated using the ground truth method by comparing the output with previous studies. The dataset was retrieved from 32 server block storage systems, encompassing metrics such as IOPS (IOs/s), service time (ms), and bandwidth (KBs/s), over a span of 1.5 years. The K-Means and DBSCAN clustering methods were applied to this dataset to form server clusters. K-Means uses a partitioning algorithm and groups servers into K clusters (e.g., K=3 for three performance levels) based on characteristic proximity. DBSCAN uses a density-based algorithm to identify flexible clusters and detect outliers. The quality of the clustering results was externally validated by comparing them to the ground truth. The performance of each algorithm was measured through accuracy levels, by examining the cluster label consistency for each server ID. Then do the cluster comparison. The outcomes of the cluster accuracy comparison demonstrate that DBSCAN is the most precise. The results of the DBSCAN clustering analysis can be used to analyze the performance of block storage servers with workloads divided into three clusters: low workload cluster, medium workload cluster, and high workload cluster. It is anticipated that this clustering will yield recommendations for workload balancing, thereby facilitating more efficient optimization of IT resource utilization.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Analisis Teknologi Informasi, Block storage, Clustering, K-Means, DBSCAN, Analisis Beban kerja Server
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Fred Erick Soaloan Sembiring
Date Deposited:	06 Aug 2025 01:18
Last Modified:	06 Aug 2025 01:18
URI:	http://repository.its.ac.id/id/eprint/127690

Actions (login required)

View Item