Storage Performance Estimation Using Processor Characteristics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing number of storage system configuration options due to software-defined storage and commodity hardware makes it impractical to test each configuration to ensure it meets desired IOPS levels, leading to potential performance discrepancies in deployed systems.
Innovation Solution
A processing platform that estimates storage system performance by utilizing characteristics of candidate processors, such as frequency, power, and benchmark performance levels, without requiring explicit testing of the configuration, using selected storage system performance models to compute and present IOPS metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If explicit testing of each storage system configuration is performed to ensure meeting IOPS requirements, then measurement precision of performance is improved, but device complexity and time consumption increase significantly
Solution Approach 1:
The patent creates a virtual copy of the storage system configuration through performance models that replicate real system behavior. Instead of physically testing each configuration, the system uses simulated environments with performance models to predict IOPS metrics, thereby maintaining measurement precision while reducing testing complexity and hardware requirements
Solution Approach 2:
The patent replaces physical mechanical testing with computational modeling and simulation. Rather than actually deploying and measuring storage systems, the system uses mathematical models and algorithms to calculate performance metrics, substituting physical measurement processes with computational analysis
2Measurement precision
If explicit testing of each storage system configuration is performed to ensure meeting IOPS requirements, then measurement precision of performance is improved, but loss of time increases significantly
Solution Approach 1:
The patent performs preliminary characterization of storage components and creates performance models in advance. By pre-establishing the relationship between component specifications and performance metrics through controlled simulations, the system can quickly predict configuration performance without time-consuming real-world testing during actual deployment
Solution Approach 2:
The patent substitutes time-consuming physical testing with rapid computational modeling. The system uses pre-characterized component data and mathematical models to instantly calculate performance metrics, reducing testing time from days or weeks to seconds while maintaining accuracy
3Adaptability or versatility
If software-defined storage with commodity hardware is used to increase configuration options, then adaptability is improved, but device complexity increases making it impractical to test all configurations
Solution Approach 1:
The patent segments the complex storage system into independent characterizable components (storage devices, processors, interconnects). Each component is individually characterized and modeled, allowing the system to handle complex configurations by combining simpler component models rather than treating the entire system as one complex unit
Solution Approach 2:
The patent introduces performance models as intermediary elements between physical configurations and performance requirements. These models act as translators that convert hardware specifications into performance predictions, mediating between the complexity of available configurations and the need for performance assurance
Data Source
AI summary
A processing platform is configured to communicate over a network with one or more client devices, and to receive a request from a given one of the client devices for a proposed configuration of a storage system. The processing platform identifies based at least in part on the received request at least one processor to be utilized in implementing the storage system, selects a particular one of a plurality of storage system performance models based at least in part on the identified processor, computes a performance metric for the storage system utilizing the selected storage system performance model and one or more characteristics of the identified processor, generates presentation output comprising: (i) the performance metric, and (ii) information characterizing at least a portion of the proposed configuration of the storage system, and delivers the presentation output to the given client device over the network.