ADAPTIVE RESOURCE SCHEDULING OPTIMIZATION

Smart overview of the Invention

The patent application introduces a method for optimizing resource scheduling in computing environments. It involves creating performance profiles for various configurations, each representing a pairing between an application and a portion of computing resources. A performance approximation formula is then computed from these profiles to evaluate each configuration's performance metric. This forms the basis of an optimization problem, which is solved to determine the optimal number of computing resources needed for application deployment, ensuring efficient resource utilization.

Background

Resource allocation in cloud computing is crucial for distributing resources like CPU, memory, and storage among applications. Traditional methods often rely on reactive algorithms that adjust resources based on workload changes, leading to inefficiencies. Modern approaches aim for dynamic allocation, scaling resources up or down according to demand while maintaining optimal performance. Artificial Intelligence (AI) technologies, including Artificial Neural Networks (ANNs) and Large Language Models (LLMs), require significant computational resources, highlighting the need for advanced resource management strategies.

Summary

The proposed method focuses on dynamic resource allocation optimization. It involves generating performance profiles, computing a performance approximation formula, and solving an optimization problem to determine the optimal resource allocation for applications. This process results in deploying applications with the optimal number of computing resources, enhancing efficiency. The method includes a computer program product and a system with program instructions for execution, providing a comprehensive solution for resource management in computing environments.

Detailed Description

Existing autoscaling methods often react to workload changes, leading to delays and inefficiencies such as over-provisioning or under-provisioning of resources. These methods may rely on single metrics like CPU usage, ignoring other factors. The patent addresses these issues by leveraging predictive analytics, adaptive scaling policies, and multi-dimensional scaling to optimize resource allocation. By considering various performance metrics and resource constraints, the proposed method aims to improve resource management efficiency and reduce operational costs.

Innovation

The patent introduces a dynamic resource optimizer, a software module that considers both hardware and software characteristics to optimize resource allocation. It uses inputs like workload demands, performance metrics, and application characteristics to generate optimized resource strategies. Unlike existing methods, it can develop optimal configurations based on specified goals such as cost minimization or power consumption reduction. This approach provides a more sophisticated and flexible solution for managing computing resources effectively.