AI Resource Provisioning for Multitenant Capacity Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional capacity planning methods for cloud services are laborious, ineffective, and expensive, often leading to inaccurate resource allocation that can result in system overloads, poor performance, or underutilized resources, particularly for large enterprise tenants.
Innovation Solution
The use of artificial intelligence models, specifically predictive models like gradient boost classifiers, to automate and enhance capacity planning by analyzing historical data and generating accurate predictions for computing resource requirements of potential tenants in multitenant computing platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional manual capacity planning methods are used, then resource allocation can be adjusted flexibly, but the process becomes laborious, time-consuming, and expensive with reduced accuracy
Solution Approach 1:
The patent replaces manual mechanical capacity planning processes with an automated machine learning system. The ML model automatically analyzes historical data, tenant characteristics, and resource usage patterns to generate capacity predictions, eliminating the need for labor-intensive manual assessment and significantly reducing planning time while improving accuracy.
Solution Approach 2:
The system enables self-service capacity planning by allowing the ML model to autonomously predict resource requirements without human intervention. The model continuously learns from historical data and automatically generates capacity recommendations, freeing administrators from manual planning tasks while maintaining high accuracy through automated analysis.
2Reliability
If traditional capacity planning methods are used, then manual control over resource allocation is maintained, but system overloads and performance issues occur due to inaccurate predictions
Solution Approach 1:
The patent replaces complex manual capacity planning procedures with a streamlined machine learning system. The ML model automatically processes multiple data sources and generates reliable capacity predictions, reducing the complexity of the planning process while improving system performance stability through accurate resource allocation forecasts.
Solution Approach 2:
The system incorporates feedback mechanisms where the ML model continuously learns from actual resource usage patterns and prediction accuracy. This feedback loop allows the model to refine its predictions over time, improving reliability while maintaining a relatively simple operational framework for capacity planning.
3Measurement precision
If specialized predictive models are used for specific tasks, then accuracy in designated areas improves, but the models cannot generalize well to tasks outside their specific domain
Solution Approach 1:
The patent segments the capacity planning problem by creating specialized ML models for different tenant types and workloads. Each model is trained on specific data patterns for particular tenant categories, achieving high accuracy for those specific segments while the overall system maintains versatility through a portfolio of specialized models that can be selected based on tenant characteristics.
Solution Approach 2:
The system achieves universality by developing a framework that can handle multiple tenant types and workload characteristics. The ML models are designed to be adaptable across different scenarios, with the ability to generalize patterns learned from specific tenant types to improve predictions for other similar workloads, balancing specialization with broad applicability.
Data Source
AI summary
A disclosed computer-implemented method may include providing, by a processor, an artificial intelligence model (“predictive model”) pre-configured to infer, via a plurality of features included in the predictive model trained using data representative of computing resource requirements of current tenants of a multitenant computing platform, computing resource requirements for potential tenants of the multitenant computing platform. The method may also include receiving, by the processor, at least one query comprising data that describes an attribute of a potential tenant of the multitenant computing platform, and determining, by the processor, based on an inference of the predictive model generated based on the at least one query, an output comprising an estimated computing resource requirement of the potential tenant. Various other methods, systems, and computer-readable media are also disclosed.