Telemetry Data Filter for Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing systems face challenges in efficiently analyzing and adapting to changing resource demands due to overwhelming amounts of telemetry data, which complicates intelligent decision-making and resource allocation, leading to potential performance degradation.
Innovation Solution
A telemetry data filter is implemented to generate a filtered data set by omitting data points where resource utilization values are subsumed by others, allowing for the identification of 'worst-case' or highest-resource-utilization time periods, enabling more efficient resource reallocation and allocation among system components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If telemetry data is collected at high frequency to capture resource usage patterns, then measurement precision is improved, but data complexity increases
Solution Approach 1:
The patent extracts only the most relevant data points from the complete telemetry dataset by identifying and removing subsumed data points. This extraction process retains the essential information needed for resource allocation decisions while eliminating redundant data, thereby reducing data complexity while preserving measurement precision for critical resource usage patterns.
Solution Approach 2:
The patent transforms the raw telemetry data by changing its structural parameters through filtering. By applying the subsumption relationship criterion (comparing resource usage across different time periods), the system reorganizes the data from a flat high-frequency dataset into a hierarchical filtered dataset that maintains precision for decision-making while reducing overall complexity.
2Loss of information
If all telemetry data points are retained for analysis, then information completeness is improved, but loss of time increases
Solution Approach 1:
The system extracts only the non-subsumed data points that contain unique information value. By identifying data points where resource usage exceeds previous maximums, the extraction process maintains information completeness for decision-making while eliminating redundant temporal data, thereby reducing analysis time without significant information loss.
Solution Approach 2:
The patent applies partial action by processing only the necessary subset of data points rather than the complete dataset. By focusing analysis on non-subsumed data points that represent critical resource usage scenarios, the system achieves sufficient information completeness for resource allocation decisions while dramatically reducing the time required for data processing and analysis.
3Adaptability or versatility
If resource allocation decisions are made frequently to adapt to changing demands, then adaptability is improved, but productivity decreases
Solution Approach 1:
The system performs preliminary filtering of telemetry data to identify non-subsumed data points before resource allocation decisions are made. This preliminary action prepares the data in advance by organizing it into a condensed format that highlights critical resource usage patterns, enabling faster and more efficient adaptability responses without the overhead of processing complete raw datasets.
Solution Approach 2:
The patent implements a dynamic data filtering mechanism that adapts to changing resource demands by continuously identifying non-subsumed data points. This dynamic approach allows the system to maintain high adaptability to varying workloads while improving productivity through optimized data processing that focuses only on relevant temporal patterns rather than static complete datasets.
Data Source
AI summary
Techniques for filtering telemetry data to allocate system resources among system components are disclosed. A system filters a data set of telemetry data prior to allocating or re-allocating system resources to system components. A filtered data set includes data points that include the highest resource-utilization values for the system components. The system compares resource-usage for each component managed by a computing machine in one time period to the resource-usage for the component in another time period. The system omits from a filtered data set any time period in which the resource-usage value for each system component is subsumed by the resource-usage values of the same system components in another time period. The system generates resource-reallocation candidate models for the computing machines in the system based on the filtered data set. The system reallocates system resources among system components using a selected resource-reallocation candidate.