OOSP: Opportunistic Optimization Scheme for Pod Deployment Enhanced with Multilayered Sensing.

Area of Science:

• Cloud Computing
• Distributed Systems
• Container Orchestration

Background:

• Kubernetes is the standard for container orchestration but its default scheduler has limitations.
• The default scheduler's focus on CPU/memory leads to suboptimal performance and resource waste in complex microservice architectures.
• Increased inter-service communication latency negatively impacts overall system performance.

Purpose of the Study:

• To propose an adaptive pod placement optimization technique for Kubernetes.
• To address the performance and resource efficiency limitations of the default Kubernetes scheduler.
• To improve application performance and resource utilization in cloud-native environments.

Main Methods:

• Developed an adaptive pod placement optimization technique using multi-tier inspection.
• Collected and analyzed multi-tier data, focusing on pod coupling and dependencies.
• Configured a Kubernetes cluster in a virtualized environment for experimental validation.

Main Results:

• The proposed method significantly outperformed the default Kubernetes scheduler.
• Achieved up to an 11.5% reduction in average response time.
• Increased requests processed per second by up to 10.04%.

Conclusions:

• The adaptive pod placement technique effectively minimizes inter-pod communication delay and enhances system-wide resource utilization.
• The method offers a more sophisticated and adaptive scheduling approach compared to traditional methods.
• This research contributes to optimizing cloud-native environments, with potential for broader application across diverse workloads and cloud platforms.