Key objectives include collecting and analyzing datasets for AI model training, designing a tailored serverless architecture optimized for AI training tasks, evaluating the performance and scalability of the architecture in real-world scenarios, and disseminating findings through research publications and presentations. This research will analyze the environmental impact and sustainability implications of adopting serverless computing for AI training. It will provide valuable insights into green computing practices and environmental conservation efforts. This research proposal outlines the following work packages to achieve the project objectives:  

WP1: Data Collection and Evaluation Metrics (6 months) 

• Month 1-3: Collect and analyze datasets for AI training, including benchmark and real-world data. 

• Month 4-6: Define evaluation metrics for assessing performance, energy efficiency, and resource utilization of serverless computing architectures in AI training. 

WP2: Serverless Computing Architecture Design for Efficient AI Training (14 months) 

• Month 7-10: Review existing serverless computing frameworks (e.g., AWS Lambda, Azure Functions) for AI model training. 

• Month 11-15: Design a serverless architecture for AI training, focusing on resource allocation and optimization. 

• Month 16-20: Implement prototype solutions for deploying AI training in a serverless environment. 

WP3: Use-case Setup (17 months) 

• Month 15-18: Develop use-case scenarios and experimental setups to emulate real-world AI training. 

• Month 19-25: Build a test bed environment with edge devices and serverless computing platforms. 

• Month 26-31: Deploy and evaluate the serverless architecture, comparing its performance against traditional server-based approaches. 

WP4: Experimentation, Evaluation and Dissemination (30 months) 

• Month 7-18: Conduct experiments to evaluate the performance and scalability of the serverless architecture.  

• Month 19-36: Analyze results to identify strengths and weaknesses, optimizing the architecture based on feedback.  

• Month 7-36: Document findings for dissemination through research publications and presentations, sharing insights with the academic and industry communities.