A Unified Experimental Framework for Evaluating and Optimizing Cloud Deployment Architectures in Smart Healthcare Systems

Abstract

In this paper, we have proposed a detailed evaluation of the deployment approaches of smart healthcare systems in the cloud computing environment, with particular emphasis on the performance comparison of Service-Oriented Architecture (SOA), containerized microservices on Amazon EC2, and serverless microservices on AWS Lambda. The proposed framework has incorporated real-time healthcare services, including patient registration, consultation, and notification, under a realistic workload model, simulating mixed read/write operations with bursty traffic patterns. The authors have also proposed the empirical evaluation of the proposed framework using a controlled experimental setup, including Linux-based EC2 instances (t3. medium), containerized Flask-based microservices, and event-driven Lambda-based microservices with concurrency limits on the number of requests. The authors have also collected the performance metrics of the proposed framework, including the average response time, throughput, CPU utilization, dropped requests, and cost per 10,000 requests. The authors have also demonstrated the performance evaluation of the proposed framework, where the serverless-based microservices (Lambda) have achieved the least average response time of 0.98 seconds, outperforming the performance of EC2-based microservices (1.05 seconds) and SOA (1.45 seconds). The authors have also demonstrated the performance evaluation of the proposed framework, where the serverless-based microservices (Lambda) have achieved the highest scalability of 2400 requests per second, outperforming the performance of EC2-based microservices (1400 requests per second) and SOA (900 requests per second). Moreover, Lambda exhibits better reliability with a lower rate of dropped requests (150), whereas SOA suffers a high degree of performance degradation with a high rate of dropped requests (1600). Cost analysis also revealed that the proposed approach using AWS Lambda is the most cost-effective approach with a cost of 0.65 USD for every 10k requests, compared to EC2 (0.85 USD) and SOA (1.20 USD). Moreover, a qualitative evaluation using expert judgment with a panel of 25 experts also supported that microservices architectures exhibit better performance in terms of increased resilience, agility in testing, and maintainability. From the analysis, it is evident that serverless microservices architectures exhibit a high degree of benefits for dynamic event-driven healthcare systems, whereas a hybrid approach using EC2 and Lambda balances both steady-state and bursty workloads for healthcare cloud systems. The study also provides a holistic approach for evaluating cloud architectures using performance metrics with decision-making for modern healthcare cloud systems.