Studi Komparatif Kinerja REST, gRPC, dan SOAP pada Strategi No-Cache, Static Cache, dan Adaptive Cache untuk Akses Data Master Sistem Distribusi FMCG

Baron, Devita Dwitama Putri (2026) Studi Komparatif Kinerja REST, gRPC, dan SOAP pada Strategi No-Cache, Static Cache, dan Adaptive Cache untuk Akses Data Master Sistem Distribusi FMCG. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025241052-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy

Abstract

Sistem distribusi Fast Moving Consumer Goods (FMCG) membutuhkan layanan perangkat lunak yang mampu menyediakan akses data master secara cepat dan stabil. Penelitian ini mengembangkan layanan Application Programming Interface (API) berbasis REST, gRPC, dan SOAP dengan tiga strategi akses data, yaitu no cache, static cache dengan Time-to-Live (TTL) tetap, dan adaptive cache berbasis Redis. Layanan API menyediakan operasi pembacaan data produk dan pelanggan yang merepresentasikan kebutuhan akses data master pada sistem distribusi FMCG. Evaluasi performa dilakukan menggunakan dataset, PostgreSQL, Redis, konfigurasi layanan, dan tingkat beban kerja yang sama pada setiap kombinasi pengujian agar perbedaan hasil terutama mencerminkan pengaruh protokol API dan strategi cache yang digunakan. Eksperimen mencakup 36 kombinasi protokol, strategi cache, dan tingkat beban kerja low, medium, high, serta stress, dengan 10 pengulangan pada setiap kombinasi sehingga menghasilkan total 360 run. Metrik evaluasi meliputi rata-rata response time, tail latency (P99), throughput, reliability error rate, dan cache hit rate. Hasil menunjukkan bahwa REST no-cache menghasilkan tail latency (P99) terendah pada seluruh tingkat beban kerja REST, dari 27,910 ms pada low hingga 73,428 ms pada stress. Pada gRPC, no-cache menghasilkan tail latency (P99) terendah pada low dan medium, sedangkan static cache lebih baik pada high dan stress dengan tail latency (P99) masing-masing 85,693 ms dan 97,108 ms. Adaptive cache memberikan manfaat paling jelas pada SOAP; pada workload stress, tail latency (P99) mencapai 98,579 ms, throughput 195,916 request/detik, dan error rate 0%. Temuan ini menunjukkan bahwa tidak terdapat satu strategi cache yang unggul secara universal. Pemilihan protokol dan strategi cache perlu mempertimbangkan karakteristik protokol, tingkat beban kerja, tail latency (P99), throughput, dan error rate secara bersamaan.
===================================================================================================================================
The Fast Moving Consumer Goods (FMCG) distribution system requires software services capable of providing fast and stable master data access. This study develops a REST, gRPC, and SOAP-based Application Programming Interface (API) service with three data access strategies: no-cache, static cache with fixed Time-to-Live (TTL), and Redis-based adaptive cache. The API service provides product and customer data read operations that represent the master data access needs of the FMCG distribution system. Performance evaluations were conducted using the same dataset, PostgreSQL, Redis, service configuration, and workload level for each test combination so that differences in results primarily reflect the influence of the API protocol and cache strategy used. The experiment included 36 combinations of protocols, cache strategies, and workload levels of low, medium, high, and stress, with 10 repetitions for each combination, resulting in a total of 360 runs. Evaluation metrics included average response time, tail latency (P99), throughput, reliability error rate, and cache hit rate. The results show that no-cache REST produces the lowest tail latency (P99) across all REST workload levels, from 27,910 ms at low to 73,428 ms at stress. On gRPC, no-cache produces the lowest tail latency (P99) at low and medium levels, while static cache performs better at high and stress levels with tail latencies (P99) of 85,693 ms and 97,108 ms, respectively. Adaptive cache provides the most obvious benefit on SOAP; on stress workloads, the tail latency (P99) reaches 98,579 ms, the throughput is 195,916 requests/second, and the error rate is 0%. These findings indicate that there is no single, universally superior cache strategy. The choice of protocol and cache strategy needs to consider the protocol characteristics, workload level, tail latency (P99), throughput, and error rate simultaneously.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Pola desain caching, Distribusi digital, FMCG, REST API, SOAP API, Kinerja sistem, Caching design pattern, Digital distribution, FMCG, REST API, SOAP API, System performance
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Devita Dwitama Putri Baron
Date Deposited:	01 Aug 2026 02:00
Last Modified:	01 Aug 2026 07:22
URI:	http://repository.its.ac.id/id/eprint/141103

Actions (login required)

View Item