UNIVERSITY CITY, Mo., April 29, 2024 (Newswire.com)
–
Tam Nguyen, CISSP, has published a new research paper in Elsevier Publications: “Holistic cold-start management in serverless computing cloud with deep learning for time series.”
The use of advanced AI, known as “Large Language Models,” has sparked interest in “smart AI agents.” These agents are intelligent programs that perform tasks on their own. Different smart AI agents can be utilized to build an AI-driven self-adaptive computing system that can automatically adjust and optimize its operations and behaviors in response to changing conditions and requirements. AI-driven self-adaptive computing systems are the holy grail to cost saving and customer satisfaction as the systems can self-adapt for minimizing resource waste, scaling up per customers’ needs, defending themselves against cyber threats, and so on.
However, these AI systems face a significant challenge: they must manage various important but conflicting tasks and goals at times. For example, in the latest type of cloud computing called “serverless computing,” resources are used very efficiently because they are allocated to small individual functions that can grow or shrink based on customer demand. But this setup struggles with the “cold-start” problem. “Cold-start” happens when there are significant delays in setting up cloud system resources for function execution. Keeping the resources ready at all times will solve the problem but is in conflict with the goal of saving resources by turning them off when not in use.
Nguyen’s paper, published in the Future Generation Computer System journal (a top-tier computer science journal), offers a solution to such problems – leveraging predictive analytics to give smart AI agents more time to work together and resolve conflicts. Nguyen also introduces a new way to test how well businesses handle the cold-start issue.
In further detail, Nguyen’s proposed novel 2-prong cold-start management policy allows feedback loops with higher-level smart AI agents while allowing lower-level cold-start optimizations. The main policy driver is the temporal convolutional network (TCN) deep learning model that can predict customers’ function requests from 5 to 15 minutes into the future. This time window is essential for all smart agents to proactively negotiate an optimal system state. Tests with real data from Microsoft and Alibaba show that this model works well.
Nguyen’s work provides practical solutions that could revolutionize how companies manage serverless cloud computing resources and improve the proactive capabilities of smart AI agents, leading to more efficient use of technologies and better service delivery to customers. Nguyen is open to further discussions at tom.nguyen@ieee.org or LinkedIn.
Source: Tam Nguyen, CISSP
Original Story: https://www.newswire.com/news/tam-nguyen-cissp-releases-new-research-paper-on-how-predictive-22307021
