Predictive Model for Data Storage Configuration Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional configuration advisory tools for data storage systems are inadequate as they fail to adapt to new failure events and cannot accurately extrapolate configuration states from unfamiliar events, relying on static heuristics derived from past expert interpretations, which limits their ability to provide optimal configurations in dynamic environments.
Innovation Solution
A predictive model is generated by a customer support center using detailed customer configuration and transaction history, allowing for real-time optimization of data storage system configurations by analyzing data such as load intensity, workload characteristics, and data access patterns, enabling the tool to respond to various system states without excessive extrapolation and supporting multiple data storage system models within a unified framework.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static heuristics are used for configuration optimization, then the tool can provide configuration advice based on historical data, but it cannot adapt to new failure events or unfamiliar configurations
Solution Approach 1:
The patent transforms the static heuristic-based advisory tool into a dynamic predictive model that continuously learns from new failure events and configuration data. The system updates its predictions in real-time based on incoming data streams from multiple storage systems, enabling it to adapt to new failure modes and configuration scenarios while maintaining reliable configuration advice.
Solution Approach 2:
The system implements feedback loops where configuration advice and failure events are collected from multiple storage systems, fed into the predictive model, and used to continuously refine future predictions. This feedback mechanism allows the system to learn from past decisions and improve its adaptability to new situations while maintaining configuration reliability.
2Adaptability or versatility
If a unified predictive model is used to combine input from multiple data storage system models, then the system achieves agnostic and responsive configuration advice, but the complexity of generating and maintaining the model increases
Solution Approach 1:
The patent creates a universal predictive model that can process input from multiple different storage system models (Clariion, VNX, Isilon) through a common interface. This unified model provides agnostic configuration advice that works across different system architectures, achieving versatility while managing complexity through standardized data collection and processing procedures.
Solution Approach 2:
The system introduces an intermediary layer that standardizes data from different storage system models before feeding it into the predictive model. This mediator component handles the complexity of model integration by providing a common data format and processing pipeline, reducing the overall system complexity while maintaining multi-model compatibility.
3Productivity
If real-time data collection from multiple customers is implemented, then the predictive model can respond to system states without excessive extrapolation, but the data management and processing requirements increase
Solution Approach 1:
The patent segments the data collection and processing system into modular components that handle data from multiple customers independently. Each storage system's data is processed through standardized pipelines that aggregate information without requiring excessive memory or computational resources. This segmentation enables real-time configuration optimization while managing data volume through distributed, incremental processing.
Data Source
AI summary
An improved technique involves generating a predictive model for data storage system configuration management. A customer support center generates such a predictive model from detailed customer configuration and transaction history. For example, a population of customers submits transaction logs to the customer support center; such transaction logs provide details as to how the customers responded to various events. The population of customers may also submit data including various statistics such as load intensity, workload characteristics, data access patterns, data change patterns, and data fingerprints to the customer support center. The customer support center then performs an analysis on the data and, from the analysis, computes values of model parameters that define a predictive model. This predictive model is configured to take in a particular state of any data storage system and produce a configuration that optimizes performance of that data storage system.