Empirical Performance and Energy Consumption Evaluation of Container Solutions on Resource Constrained IoT Gateways

Area of Science:

• Computer Science
• Embedded Systems
• Network Engineering

Background:

• Containers offer lightweight software deployment for edge and Internet of Things (IoT) devices, reducing latency and energy consumption.
• Existing research on containerized services for IoT often overlooks the constraints of limited energy and computing resources, hindering scalability.
• IoT devices increasingly require data analytics and intelligence under strict latency and energy budgets.

Purpose of the Study:

• To establish guidelines and identify critical factors for deploying containerized services on resource-constrained IoT devices.
• To compare the performance of Docker Swarm and Kubernetes for container orchestration in an IoT context.
• To evaluate the impact of container deployment on key performance indicators relevant to IoT environments.

Main Methods:

• Two container orchestration tools, Docker Swarm and Kubernetes, were tested on a baseline IoT gateway testbed.
• Experiments involved deploying Deep Learning-driven data analytics and Intrusion Detection System services.
• Key metrics evaluated included container creation time, CPU utilization, memory usage under varying traffic loads, and energy consumption.

Main Results:

• Container creation time and memory usage were identified as decisive factors for microservice architecture on IoT devices.
• Performance differences between Docker Swarm and Kubernetes were observed in terms of resource utilization and deployment efficiency.
• Resource constraints significantly impact the scalability and concurrent deployment of containerized services on IoT gateways.

Conclusions:

• Container creation time and memory usage are critical parameters for successful containerized microservice deployments on resource-constrained IoT devices.
• Guidelines are needed to optimize container orchestration strategies for edge computing and IoT environments.
• Further research should focus on resource-aware scheduling and management for containerized workloads in IoT ecosystems.