A fuzzy TOPSIS approach to dynamic routing under uncertainty with an application to microservices

Dynamic routing in complicated network settings means weighing a number of conflicting factors while dealing with built-in measurement uncertainty and network jitter. Conventional deterministic load-balancing algorithms frequently fail to uphold Quality of Service (QoS) because they cannot accurately represent this uncertainty. This paper presents a generalized fuzzy multi-criteria decision-making framework that combines triangular fuzzy numbers with the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to address dynamic network uncertainty. To substantiate the proposed framework, we offer an extensive case study on microservice routing within a service mesh architecture. The suggested method considers service latency, load, and reliability as imprecise variables. We use the Analytic Hierarchy Process (AHP) to figure out the weights of the criteria and make a complete ranking system for choosing service instances. A publicly available sample of the Google Cluster Trace 2019 workload shows that our Fuzzy TOPSIS-based router works much better than standard algorithms, with an average latency of 107.97 ms (a 62.2% reduction compared to Round Robin). The framework also has a low SLA violation rate of 3.21% even when there are many different types of users at the same time. This shows that it is more robust than Weighted Round Robin, which has a violation rate of 16.44%.