Automatic Rule Learning for Cloud Solution Design Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In shared resource environments like cloud computing, designing solutions for multiple virtual machines with complex configurations is time-consuming and requires experienced architects, leading to inefficiencies and long design cycles, as each new customer's solution must be created from scratch.
Innovation Solution
Automatically learning shared resource environment solution design rules from a collection of requirement-solution pairs, where a processor iteratively generates candidate design rules, filters them based on evidence scores, and optimizes the rule set by merging design rules, enabling semi-automatic solution generation and knowledge transfer among architects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual solution design by experienced architects is used, then solution quality and customer requirements satisfaction are improved, but design time and productivity deteriorate
Solution Approach 1:
The patent creates a digital twin or virtual model of the manual design process by training an AI model on historical design data. The model learns from existing high-quality solutions created by experienced architects and generates new solutions that replicate their expertise, eliminating the need for manual copying of design patterns while maintaining solution quality.
Solution Approach 2:
The patent replaces the mechanical system of manual architectural design with an automated AI-based system. The neural network model processes customer requirements and generates solution configurations automatically, substituting human cognitive processes with computational algorithms that can operate at much higher speeds while maintaining or improving solution quality through consistent application of learned design principles.
2Adaptability or versatility
If case-by-case manual design is used, then customized solutions meeting specific requirements are improved, but scalability and ability to serve new customers deteriorate
Solution Approach 1:
The patent creates a universal design system that can handle multiple different customer requirements and solution types through a single AI model. The model is trained on diverse historical design data covering various scenarios, enabling it to adapt to different customer needs while maintaining consistent quality standards, thus providing multi-functional capability that scales across numerous customers simultaneously.
Solution Approach 2:
The patent utilizes parameter changes in the AI model's processing to achieve scalability. By adjusting input parameters representing different customer requirements and using the model's learned relationships between parameters, the system can generate customized solutions for any number of customers by simply changing the input parameters rather than requiring manual redesign for each case.
3Measurement precision
If experienced architects manually design solutions, then expertise and solution accuracy are improved, but knowledge transfer and training of new architects deteriorate
Solution Approach 1:
The patent copies the implicit knowledge and expertise of experienced architects by training the AI model on their historical design work. The model learns the patterns, principles, and decision-making processes that experts use, capturing this knowledge in a digital form that can be consistently applied and transferred to new architects through the system's generated solutions and explanations.
Solution Approach 2:
The patent implements feedback mechanisms where the AI model provides explanations and rationales for its design decisions, allowing new architects to learn from the system's output. The model can also be refined based on feedback from experienced architects reviewing its solutions, creating a continuous learning loop that preserves and transfers expertise while improving solution accuracy over time.
Data Source
AI summary
One embodiment provides automatically learning shared resource environment solution design rules from a collection of requirement-solution pairs including obtaining requirement-solution pairs for a shared resource environment from a data store. A processor iteratively generates a candidate design rule set from each requirement-solution pair. Each generating iteration uses an input including the candidate design rule set output from a previous generating iteration. Evidence scores of each candidate design rule are calculated and candidate design rules having higher evidence score than an evidence score threshold are retained to obtain a learned design rule set. Candidate rules of a next iteration are constructed based on an addition of new attributes to rules of the learned design rule set.