Experimental Analysis of the Application of Serverless Computing to IoT Platforms

Area of Science:

• Computer Science
• Cloud Computing
• Internet of Things

Background:

• Serverless computing, particularly Function-as-a-Service (FaaS), is a popular application deployment model.
• FaaS platforms offer warm start (perceived constant availability) and cold start (resource-saving container destruction) deployment modes.

Purpose of the Study:

• To evaluate the suitability of warm start versus cold start FaaS deployment for Internet of Things (IoT) applications.
• To analyze resource consumption and latency on a low-resource testbed simulating an edge node.

Main Methods:

• Implementation of an IoT serverless platform using OpenFaaS on a resource-constrained testbed.
• Experimental performance analysis comparing warm start and cold start modes under varying conditions.
• Configuration of OpenFaaS cold start with precise inactivity time settings.

Main Results:

• Cold start mode demonstrates potential for significant resource savings on edge nodes.
• The effectiveness of cold start is contingent on the data transmission period relative to container shutdown time.
• Latency and resource consumption were evaluated for both deployment modes.

Conclusions:

• Cold start FaaS deployment can be advantageous for resource-constrained IoT edge nodes.
• Optimal use of cold start requires careful tuning of inactivity timers to match application data transmission patterns.
• Serverless FaaS is a viable option for IoT, with cold start offering efficiency benefits when properly configured.