Architectural Optimization Of A Lightweight Kubernetes Microservices Framework For The Toba Herbal Information System Under Extreme Resource Constraints

Abstract

This study designs, implements, and evaluates a lightweight, decoupled virtualization architecture utilizing K3s Kubernetes and Docker containers tailored for the Toba Herbal Information (THI). The system architecture isolates the web application logic, stateful relational databases, and high-velocity user connection sessions into three independent, resource-constrained pod tiers. Operating under a strict single-node hardware envelope of 4 CPU cores and 2 GB RAM, the framework enforces rigorous memory quotas and integrates an in-memory Redis caching buffer. Performance resiliency was rigorously evaluated via Apache JMeter against a baseline monolithic virtual machine (VM) utilizing a peak load of 1,000 concurrent healthcare users, complemented by an intentional Chaos Engineering failure injection. Empirical benchmarks demonstrate that the proposed 3-pod K3s framework successfully accommodated peak traffic with a stable 95th percentile (P_95) latency under 340ms and zero Out-Of-Memory (OOM) disruptions, safely capping RAM saturation at 87%. Conversely, the monolithic setup suffered a catastrophic kernel freeze at 400 concurrent users. Following a forced application pod termination, the K3s control plane achieved an exceptional Mean Time to Recover (MTTR) of 3.8 seconds with 100% transactional data integrity. This paper establishes a repeatable, cost-effective architectural blueprint proving that defensive microservices orchestration can successfully overcome physical hardware barriers to deliver enterprise-grade digital health platforms in resource-constrained rural regions.