Optimizing Wso2 Api Manager for Improved Graphql Query Performance in a Distributed Setup

Abstract

In the rapidly evolving landscape of API management, GraphQL has emerged as a powerful query language that enables efficient data retrieval. However, optimizing GraphQL query performance within the context of distributed WSO2 API Manager remains a significant challenge. This research aimed to address this challenge by investigating the performance bottlenecks of GraphQL queries in WSO2 API Manager and proposing a comprehensive optimization model. The study begins by conducting a thorough literature review to identify existing performance challenges and potential optimization techniques. Through empirical testing and analysis, a baseline performance profile is established, highlighting key areas for improvement. The proposed optimization model encompasses three main components: JVM optimization, query splitting algorithm, and caching mechanisms. JVM optimization focuses on tuning parameters such as heap size and garbage collection settings to enhance resource utilization. The query splitting algorithm which exploits parallel execution to enhance performance was applied. Caching mechanisms, implemented using Redis, enable efficient storage and retrieval of frequently accessed data, minimizing redundant database queries. The effectiveness of the optimization model was evaluated using the M-PESA Payment API as a real-world test case. Experiments were conducted under various concurrency levels, measuring key performance metrics such as response time, throughput, and error rate. The optimized performance was compared against the baseline, revealing significant improvements. The model achieves an average response time reduction of 4.62% and throughput growth of 5.89%, demonstrating its efficacy in enhancing GraphQL query performance. This study evaluates three complementary techniques in a distributed WSO2 API Manager: (i) server-level optimization (OS TCP tuning and JVM G1GC with fixed heap), (ii) query splitting for parallel sub‑resolution of GraphQL fields, and (iii) Redis-based distributed caching for tokens and hot query results.