Decentralized Probability Active Sampling for System Configuration Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current computing systems face challenges in determining optimal configuration settings due to their growing complexity and heterogeneity, leading to non-optimal performance when system components from different vendors interact with default settings, which ignore interdependencies.

Innovation Solution

A decentralized probability-based active sampling (DPAS) approach is employed to automatically identify optimal configurations by treating the system as a black-box and using sampling-based optimization, incorporating multiple performance dimensions, and dynamically updating probabilities based on reward-penalty strategies to efficiently explore the configuration space.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If default configuration settings are used from vendors, then system deployment is conservative and simple, but system performance is non-optimal due to ignoring interdependencies among components

Engineering Contradiction:

Improvesystem deployment simplicityVSAvoidsystem performance

Core Design Contradiction:

Ease of manufactureVSProductivity

Solution Approach 1:

The system performs self-configuration by automatically discovering optimal settings through iterative sampling and performance evaluation, eliminating the need for manual expert tuning while achieving better performance than vendor defaults

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The configuration optimization process uses feedback from performance measurements to iteratively improve settings, where system performance is evaluated under different configurations and results are used to guide subsequent sampling decisions

Inventive Principle:

Principle #23Feedback

2Ease of operation

If manual configuration tuning is performed by human operators, then configuration expertise can be applied, but it is difficult to find optimal settings for large complex systems with thousands of parameters

Engineering Contradiction:

Improveconfiguration expertise utilizationVSAvoidsystem parameter complexity

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The patent replaces manual human configuration tuning with an automated computational system that uses probabilistic sampling and performance-based evaluation to discover optimal settings, substituting human expertise with algorithmic optimization

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The configuration space is segmented into multiple dimensions corresponding to different system parameters, allowing the optimization process to systematically explore and tune each parameter independently or in combination with others

Inventive Principle:

Principle #1Segmentation

3Productivity

If conventional optimization algorithms are used to search for optimal configuration, then systematic search can be performed, but the algorithms struggle with unknown non-convex functions with multiple local maxima

Engineering Contradiction:

Improvesystematic search capabilityVSAvoidoptimization convergence

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The optimization approach uses dynamic probabilistic sampling where the sampling distribution is continuously adapted based on observed performance, allowing the system to dynamically shift exploration focus from high-performing regions to potentially better regions, avoiding static algorithmic limitations

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameters of the sampling distribution itself during optimization, adjusting which configuration values are sampled based on performance feedback, rather than using fixed optimization algorithm parameters

Inventive Principle:

Principle #35Parameter changes

4Adaptability or versatility

If the number of configuration parameters increases to handle system scalability, then system capability improves, but the complexity of finding optimal configuration increases exponentially

Engineering Contradiction:

Improvesystem scalabilityVSAvoidconfiguration search complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system performs partial optimization by focusing sampling efforts on the most influential parameters identified through performance feedback, rather than exhaustively searching all parameter combinations, achieving good results with limited sampling in high-dimensional spaces

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS8095486B2Discovering optimal system configurations using decentralized probability based active sampling

Publication Date: 2012.01.10 NEC CORP

AI summary

A system and method for optimizing system performance includes applying sampling based optimization to identify optimal configurations of a computing system by selecting a number of configuration samples and evaluating system performance based on the samples. Based on feedback of evaluated samples, a location of an optimal configuration is inferred. Additional samples are generated towards the location of the inferred optimal configuration to further optimize a system configuration.