Container Orchestration Patterns for Optimizing Resource Use
Diogo Maia, Filipe Correia, André Restivo, Paulo Queiroz
TL;DR
This paper addresses the challenge of unclear orchestration practices in container-based service architectures by formalizing three design patterns: Preemptive Scheduling, Service Balancing, and Garbage Collection. The authors derive these patterns from literature and tooling analysis, demonstrating how they address deployment-time and run-time orchestration tasks. The contributions include a pattern map linking the new patterns to established ones, detailed pattern definitions and implementation guidance, and discussions of consequences, known uses, and related patterns. The work aims to broaden adoption of consistent, pattern-based orchestration in cloud-native environments and suggests directions for expanding the pattern language to cover more operational concerns.
Abstract
Service-based architectures provide substantial benefits, yet service orchestration remains a challenge, particularly for newcomers. While various resources on orchestration techniques exist, they often lack clarity and standardization, making best practices difficult to implement and limiting their adoption within the software industry. To address this gap, we analyzed existing literature and tools to identify common orchestration practices. Based on our findings, we define three key orchestration resource optimization patterns: {\sc Preemptive Scheduling}, {\sc Service Balancing}, and {\sc Garbage Collection}. {\sc Preemptive Scheduling} allows the allocation of sufficient resources for services of higher priority in stressful situations, while {\sc Service Balancing} enables a restructuring of the nodes to allow better resource usage. To end, {\sc Garbage Collection} creates cleanup mechanisms to better understand the system's resource usage and optimize it. These patterns serve as foundational elements for improving orchestration practices and fostering broader adoption in service-based architectures.