<![CDATA[The demand for reliable, real-time online attendance systems capable of handling large-scale users continues to increase alongside the widespread adoption of Learning Management Systems (LMS) in higher education and online training. Conventional attendance methods based on manual input or simple authentication mechanisms suffer from weaknesses such as susceptibility to fraud, limited automation, and degraded performance under high workloads. Face recognition has emerged as a promising alternative, as it enables automatic and non-intrusive user identity verification. However, most face-based attendance systems still rely on centralized server-side processing, which leads to high latency and limited scalability. This study aims to design and evaluate an online attendance architecture that integrates WebAssembly and Microservices by separating computational workloads between the client and server. The Design Science Research method is employed to develop a web-based face attendance application as the research artifact, in which face detection and feature extraction are executed entirely on the client side using OpenCV.js compiled to WebAssembly, while authentication, attendance recording, and session management are handled by a Microservices-based backend. The evaluation includes face recognition accuracy testing, end-to-end latency measurement, and system throughput analysis. Experimental results demonstrate that the proposed architecture reduces attendance latency by approximately 72 percent compared to a monolithic server-side processing approach, while simultaneously increasing request handling capacity without compromising accuracy. These findings indicate that the integration of WebAssembly and Microservices constitutes an effective architectural solution for real-time biometric attendance systems. ABSTRAKKebutuhan akan sistem presensi daring yang andal, real-time, dan mampu menangani skala pengguna besar terus meningkat seiring dengan meluasnya penggunaan Learning Management System (LMS) dalam pendidikan tinggi dan pelatihan daring. Metode presensi konvensional berbasis input manual maupun autentikasi sederhana memiliki kelemahan berupa potensi kecurangan, keterbatasan otomatisasi, serta performa yang menurun pada kondisi beban tinggi. Pengenalan wajah menjadi solusi alternatif yang menjanjikan karena mampu memverifikasi identitas pengguna secara otomatis dan non-intrusif. Namun, sebagian besar sistem presensi berbasis wajah masih bergantung pada pemrosesan terpusat di sisi server, yang mengakibatkan latensi tinggi dan keterbatasan skalabilitas. Penelitian ini bertujuan merancang dan mengevaluasi arsitektur presensi daring berbasis integrasi WebAssembly dan Microservices dengan pendekatan pemisahan beban komputasi antara klien dan server. Metode Design Science Research digunakan untuk mengembangkan artefak berupa aplikasi presensi wajah berbasis web, di mana proses deteksi dan ekstraksi fitur wajah dijalankan sepenuhnya di sisi klien menggunakan OpenCV.js yang dikompilasi ke WebAssembly, sedangkan autentikasi, pencatatan presensi, dan manajemen sesi ditangani oleh backend berbasis Microservices. Evaluasi dilakukan melalui pengujian akurasi pengenalan wajah, pengukuran latensi end-to-end, dan analisis throughput sistem. Hasil pengujian menunjukkan bahwa arsitektur yang diusulkan mampu menurunkan latensi presensi sekitar 72 persen dibandingkan pendekatan monolitik berbasis pemrosesan server, sekaligus meningkatkan kapasitas penanganan permintaan tanpa mengorbankan tingkat akurasi. Temuan ini menunjukkan bahwa integrasi WebAssembly dan Microservices merupakan solusi arsitektural yang efektif untuk sistem presensi biometrik real-time.]]>
Copyrights © 2026
