Latency-Sensitive Function Placement among Heterogeneous Nodes in Serverless Computing

Area of Science:

• Computer Science
• Artificial Intelligence
• Smart City Technologies

Background:

• Function as a Service (FaaS) provides serverless computing, simplifying infrastructure management for developers.
• FaaS integration with the Internet of Things (IoT) enables event-driven actions and real-time computations, crucial for smart cities.
• Optimizing function placement in FaaS is critical for meeting performance requirements like deadlines and efficient resource use.

Purpose of the Study:

• To develop and evaluate a likelihood-based adaptive machine learning model for optimal FaaS function placement.
• To enhance resource utilization and ensure deadline compliance in FaaS deployments within smart city contexts.
• To address the challenges of network latency and computational demands in distributed FaaS environments.

Main Methods:

• Employed an adaptive machine learning model utilizing XGBoost regressor for execution time estimation and decision tree regressor for network latency prediction.
• Incorporated factors such as network delay, arrival computation, and resource emphasis into the machine learning model for placement decisions.
• Utilized Docker containers for replication, focusing on serverless node types, function location, deadlines, and edge-cloud topology.

Main Results:

• The proposed machine learning model effectively aids in selecting the optimal placement for FaaS functions.
• Effective resource utilization was demonstrated to directly correlate with enhanced deadline compliance.
• The research validates the benefits of FaaS in smart city infrastructure through optimized placement strategies.

Conclusions:

• The adaptive machine learning model provides an effective solution for optimizing FaaS function placement in smart cities.
• Efficient resource management and meeting strict deadlines are achievable through intelligent FaaS deployment strategies.
• This research contributes to the advancement of serverless computing for scalable and responsive smart city applications.