Rancang Bangun Backend Sistem Pemantauan Baku Mutu Pome Berbasis Container Dengan Pendekatan Devops

Reba, Alvin Vincent Oswald (2026) Rancang Bangun Backend Sistem Pemantauan Baku Mutu Pome Berbasis Container Dengan Pendekatan Devops. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221201-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Palm Oil Mill Effluent (POME)bervolumebesardanharusmemenuhibakumutulingkungan, termasuk saat diolah dengan mikroalga yang memerlukan kualitas air terjaga. Pemantauan manual berkala menyebabkan perubahan kualitas dan pelampauan baku mutu terlambat terdeteksi. Penelitian ini bertujuan merancang, membangun, dan mengevaluasi backend sistem pemantauan baku mutu POME berbasis Internet of Things (IoT) dan kontainer dengan arsitektur microservices serta pendekatan DevOps (Development and Operations). Metode Research and Development diterapkan melalui strategi Iterative and Incremental Development. Sistem dibangun bertahap dan mencakup akuisisi data multiprotokol, autentikasi Single Sign-On (SSO), Role-Based Access Control (RBAC), isolasi data antar-tenant, serta integrasi flow service dan layanan inferensi Machine Learning (ML) berbasis gRPC pada Kuber netes. Backend Go/Gin menggunakan basis data hibrida PostgreSQL-InfluxDB dan Kong API Gateway, sedangkan pipeline Continuous Integration/Continuous Delivery (CI/CD) mengatur build, pengujian, dan rilis. Pengujian unit dan fungsional pada empat layanan utama berjalan tanpa kegagalan, dengan cakupan kode inti 95,5%. Pengukuran DevOps Research and Assessment (DORA) selama 365 hari pada cakupan platform mencatat 271 job deployment berhasil dari 293 upaya, frekuensi deployment 5,20 kali per minggu, dan Lead Time for Changes 36,51 menit. Sebanyak 22 job rollout gagal (7,51%) merupakan indikator proxy, bukan Change Failure Rate (CFR) produksi. Karena tidak ada kandidat insiden dengan validated=true, CFR, Mean Time to Restore, dan Failed Deployment Recovery Time (FDRT) produksi tidak tersedia. Eksperimen sandbox terpisah menghasilkan median controlled Failed-Deployment Recovery Time (cFDRT) khusus penelitian sebesar 97,97 detik, bukan FDRT DORA produksi. Perbandingan terkontrol atas lima protokol akuisisi, yaitu HTTP (Hypertext Transfer Protocol), WebSocket, MQTT (Message Queuing Telemetry Transport), CoAP (Constrained Application Protocol), dan FCP (Fast Control Protocol), menemukan bahwa tidak ada protokol yang unggul pada semua beban. HTTP dan FCP hampir setara serta paling stabil untuk telemetri sensor, sedangkan HTTP paling cepat untuk transfer citra. Hasil tersebut menunjukkan bahwa pengembangan bertahap, kontainerisasi, CI/CD, dan Kubernetes meningkatkan konsistensi deployment serta menyediakan dasar evaluasi operasional yang terukur bagi pemantauan baku mutu POME secara real-time.
======================================================================================================================================
Large volumes of Palm Oil Mill Effluent (POME) must meet environmental quality standards, including when treated with microalgae that require controlled water quality. Periodic manual monitoring delays the detection of quality changes and standard exceedances. This study aims to design, build, and evaluate the backend of an Internet of Things (IoT) and container-based POME quality-monitoring system with a microservices architecture and a DevOps (Development and Operations) approach. The Research and Development method was applied through an Iterative and Incremental Development strategy. The system was built in stages and supports multiprotocol data acquisition, Single Sign-On (SSO) authentication, Role-Based Access Control (RBAC), inter-tenant data isolation, and integration with a flow service and a gRPC-based Machine Learning (ML) inference service on Kubernetes. The Go/Gin backend uses a hybrid PostgreSQL-InfluxDB database and Kong API Gateway, while a Continuous Integration/Continuous Delivery (CI/CD) pipeline handles building, testing, and release. Unit and functional tests on the four main services passed without failure, with core code coverage of 95.5%. DevOps Research and Assessment (DORA) measurements over 365 days at the platform scope recorded 271 successful deployment jobs from 293 attempts, a deployment frequency of 5.20 per week, and a Lead Time for Changes of 36.51 minutes. The 22 failed rollout jobs (7.51%) are a proxy indicator, not a production Change Failure Rate (CFR). Because no incident candidate had an explicit validated=true flag, the production CFR, Mean Time to Restore, and Failed Deployment Recovery Time (FDRT) are unavailable. A separate sandbox experiment yielded a study-specific median controlled Failed-Deployment Recovery Time (cFDRT) of 97.97 seconds, not production DORA FDRT. A controlled comparison of five acquisition protocols, namely HTTP (Hypertext Transfer Protocol), WebSocket, MQTT (Message Queuing Telemetry Transport), CoAP (Constrained Application Protocol), and FCP (Fast Control Protocol), found that no protocol excels under every workload. HTTP and FCP were nearly equal and the most stable for sensor telemetry, while HTTP was the fastest for image transfer. These results indicate that incremental development, containerization, CI/CD, and Kubernetes improve deployment consistency and provide a measurable operational basis for real-time monitoring of POME quality standards.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	DevOps, DORA, IoT, Kubernetes, POME, real-time monitoring.
Subjects:	T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Alvin Vincent Oswald Reba
Date Deposited:	24 Jul 2026 04:13
Last Modified:	24 Jul 2026 04:13
URI:	http://repository.its.ac.id/id/eprint/137790

Actions (login required)

View Item