Cloud Resource Spend Clustering for Actionable Cost Anomalies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud cost optimization tools fail to effectively identify systemic spend anomalies and provide actionable recommendations due to their inability to analyze cloud resources in a holistic manner, leading to unpredictable and chaotic cloud expenditures.
Innovation Solution
A method and system that identifies homogeneous resources, creates two-dimensional clusters using k-means clustering, computes effective prices, and uses decision trees and Interesting Subset Discovery (ISD) to detect spend anomalies, iteratively relax attribute constraints, and expand the search space for optimization recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing tools analyze cloud resources in isolation to detect spend anomalies, then detection capability is improved, but the ability to capture systemic impact and provide prioritized recommendations deteriorates
Solution Approach 1:
The patent segments cloud resources into homogeneous sets based on multiple attributes (service type, resource type, deployment location, cost center, etc.), then analyzes each segment separately before aggregating results. This allows detailed anomaly detection within segments while capturing systemic impacts across segments, resolving the contradiction between detection precision and systemic analysis capability.
Solution Approach 2:
The patent transitions from analyzing resources in isolation (one-dimensional) to analyzing them across multiple dimensions simultaneously (multi-dimensional). By introducing additional dimensions such as attribute constraints, homogeneous set relationships, and systemic impact assessments, the system achieves both precise anomaly detection and comprehensive systemic understanding without excessive complexity.
2Loss of energy
If existing tools detect spend leakages, then cost optimization is improved, but actionable recommendations and prioritization deteriorate
Solution Approach 1:
The patent implements a feedback mechanism that continuously refines anomaly detection and recommendation generation based on analyzed data. The system provides actionable recommendations by feedback loops that validate detected anomalies against business context, ensuring recommendations are not only detected but also prioritized and actionable for users.
Solution Approach 2:
The patent dynamically adjusts detection parameters and analysis depth based on the complexity of homogeneous sets and the impact of identified anomalies. By changing parameters such as the number of attributes considered, clustering granularity, and recommendation specificity based on context, the system maintains ease of operation while providing targeted, actionable recommendations for cost optimization.
3Adaptability or versatility
If cloud resources are managed with high flexibility and variability, then adaptability is improved, but spend predictability and budget control deteriorate
Solution Approach 1:
The patent performs preliminary analysis by identifying homogeneous sets and baseline prices before actual spend occurs. By establishing expected cost patterns in advance through clustering and anomaly detection, the system enables predictability while maintaining the flexibility to accommodate varying cloud resource needs through continuous monitoring and adaptive recommendations.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Existing tools detect abnormal spends but fail to capture their systemic impact as these tools analyze resources in isolation and further fail to offer actionable recommendations. The present disclosure identifies one or more set of homogeneous resources from one or more set of resources. One or more two-dimensional clusters are created between dimensions of spend and dimensions of quantity. An effective price for each of one or more two-dimensional clusters is created and a baseline price is identified. One or more spend anomalies are identified based on comparison of associated effective prices of one or more two-dimensional clusters and identified baseline price. One or more attribute constraints are identified which when relaxed provide maximum reduction in defined baseline price to rectify identified one or more anomalies. Expands a search space to generate one or more recommendations within a new search space with relaxed one or more attribute constraints.