COMPARATIVE ANALYSIS OF SERVICE ORCHESTRATION TOOLS IN MICROSERVICES ARCHITECTURE

Abstract

The article examines a comparison of two orchestration tools—Docker Swarm and Kubernetes. The choice of these technologies is driven by their widespread use in practical projects, allowing for results that have direct practical value for the IT industry. A literature review on this topic was conducted, revealing that existing research pays little attention to analyzing the impact of architectural decisions on availability parameters. A comparative analysis was performed, and three of the most common types of failures were selected for system failure simulation: the complete shutdown of a node, network communication loss between nodes, and a sudden traffic spike. Each scenario was reproduced ten times. The criteria for assessing system availability were chosen as the average recovery time after failure and the percentage of service availability. The research results showed that Kubernetes has a significantly higher level of availability compared to Docker Swarm in all tested scenarios. In the case of node failure simulation, Kubernetes demonstrated an availability level of 99.95%, whereas Docker Swarm provided only 95.20%. When modeling network delays, Kubernetes maintained an availability of 99.85%, which is 3.10% higher than Docker Swarm (96.75%). During the simulation of sudden peak loads (Load Spike), Kubernetes exhibited resilience to fluctuating workloads, ensuring 99.92% availability, while Docker Swarm achieved only 95.10%. In the event of a complete node shutdown, Kubernetes restored operation within an average of 15 seconds, whereas Docker Swarm required 42 seconds. The obtained results have direct practical value for developers, software architects, and DevOps specialists. The study demonstrates that Kubernetes is a superior tool for ensuring high service availability in a microservices architecture.