Storage Object Temperature Forecasting via Sub-Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large storage systems, machine-learning-based approaches for forecasting the future activity of storage objects are computationally expensive due to high memory footprint and CPU overhead, making it prohibitive to accurately predict the temperature of millions of objects.
Innovation Solution
Divide storage objects into sub-objects and use a simple moving average to determine their temperature, then project the temperature of the main object onto its sub-objects, allowing for less computationally intensive processing and more accurate forecasting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine-learning-based approaches are used to forecast storage object temperature, then forecasting accuracy is improved, but computational cost (memory footprint and CPU overhead) increases
Solution Approach 1:
The patent divides storage objects into smaller sub-objects (e.g., files into sub-files, extents into sub-extents). Instead of applying complex machine learning to all storage objects, the system applies ML only to selected parent objects or uses simplified methods for sub-objects, thereby reducing overall computational cost while maintaining forecasting accuracy through hierarchical aggregation of sub-object temperatures to parent object temperatures.
2Measurement precision
If machine-learning-based approaches are applied to all storage objects, then forecasting accuracy is improved, but processing time increases
Solution Approach 1:
By segmenting storage objects into sub-objects, the system can process temperatures at multiple granularities. Sub-object temperatures are computed using simple methods (e.g., moving averages) quickly, then aggregated to parent object temperatures. This hierarchical approach significantly reduces processing time compared to applying ML to all objects individually, while still providing accurate forecasts through the aggregation mechanism.
Solution Approach 2:
The system applies machine learning selectively rather than universally. It identifies which parent storage objects benefit most from ML-based forecasting and applies the computationally intensive method only to those cases. For other objects, simpler forecasting methods are used, reducing overall processing time while maintaining accuracy where it matters most.
3Device complexity
If simple statistical methods are used to forecast storage object temperature, then computational cost is reduced, but forecasting accuracy deteriorates
Solution Approach 1:
The patent uses a hierarchical segmentation approach where sub-objects are processed with simple statistical methods (low computational cost), and parent objects are processed with machine learning methods (higher accuracy). The aggregation of sub-object temperatures to parent object temperatures combines the benefits of both approaches, achieving good forecasting accuracy without the full computational cost of applying ML to every object.
Solution Approach 2:
The system introduces an intermediary aggregation layer between sub-objects and parent objects. Sub-object temperatures computed by simple statistical methods are aggregated to form parent object temperatures, which then serve as inputs to machine learning models. This intermediary aggregation mechanism bridges the gap between simple and complex methods, achieving balanced accuracy and computational cost.
Data Source
AI summary
A method, computer program product, and computing system for forecasting a temperature of a storage object of a storage system using a machine learning model. The storage object may be divided into a plurality of storage sub-objects. A temperature may be determined for each storage sub-object using a simple moving average. A portion of the temperature of the storage object may be projected onto the temperature of each of the plurality of storage sub-objects based upon, at least in part, the temperature determined for each storage sub-object and the temperature determined for each storage object.