Clustering Routines for Computing Resource Cost Metric Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud management platforms face challenges in generating accurate cost metrics for data centers due to the need for manual input of purchase costs for numerous servers, which is cumbersome and often results in inaccurate estimates, as standard reference libraries fail to account for discounts and other factors.
Innovation Solution
Implementing clustering routines, such as Gaussian-means and k-means, to group computing resources with similar configuration parameters, allowing for the extrapolation of cost metrics from a minimalistic list of inputs, reducing the number of required inputs while maintaining high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual input of purchase costs for numerous servers is required, then cost metric accuracy can be improved, but the ease of operation deteriorates due to the cumbersome process
Solution Approach 1:
The patent segments the large set of servers into smaller clusters based on configuration parameters. Instead of requiring manual input for each individual server, the system groups servers with similar configurations into clusters and only requires manual cost input for one representative server per cluster. This segmentation reduces the number of required inputs from hundreds or thousands of servers to just a few cluster representatives, thereby improving ease of operation while maintaining cost metric accuracy through the clustering extrapolation process.
2Ease of operation
If the number of input values is reduced from hundreds to a few, then the ease of operation is improved, but the measurement precision of cost metrics may worsen
Solution Approach 1:
The system performs preliminary clustering analysis based on configuration parameters before cost metric calculation. By pre-grouping servers into clusters using their configuration characteristics (CPU, memory, storage, etc.), the system establishes relationships between servers that allow cost extrapolation. This preliminary action enables the system to maintain measurement precision by leveraging the configuration-based clustering structure, even when manual cost inputs are reduced to just a few representative values per cluster.
Solution Approach 2:
The patent uses copying by extrapolating cost metrics from representative servers to entire clusters of servers with similar configurations. Once a manual cost input is obtained for one server in a cluster, the system copies this cost metric to all other servers in the same cluster based on their configuration similarity. This copying mechanism allows the system to maintain accurate cost metrics across hundreds or thousands of servers while requiring manual input for only a handful of representative servers.
3Ease of operation
If standard reference libraries are used for cost estimation, then the ease of operation is improved, but the measurement precision deteriorates due to failure to account for discounts and other factors
Solution Approach 1:
The system implements self-service by automatically performing clustering analysis and cost extrapolation without requiring users to manually configure reference libraries or input data for every server. The system autonomously groups servers by configuration, identifies representative servers for each cluster, and extrapolates costs across clusters. This self-service approach maintains ease of operation while improving measurement precision by capturing actual cost variations across different server configurations and applying appropriate discounts or adjustments based on the clustering relationships.
Data Source
AI summary
Various examples are disclosed for using clustering routines to extrapolate metrics to other computing resources in a cluster. One or more computing devices can classify computing resources, such as servers, based on various characteristics of the computing resources. For each class of computing resource, a clustering routine can be applied to generate clusters of the computing resources. A minimal number of metrics required to be obtained from an end user can be determined as a function of a number of the clusters. If one or more of the metrics are obtained from the end user, the metrics can be extrapolated to other computing resources in the same cluster.