AI-Based Cache Distribution for Edge Data Centers

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Edge computing platforms face challenges in managing customer-specific data across diverse customer sites, requiring decisions on data caching, caching frequency, and communication with core data centers, which is complex due to unique customer requirements and compliance issues, especially with traditional machine learning methods that do not account for heterogeneous products and usage patterns.

Innovation Solution

Implementing a cohesive and distributed machine learning approach with artificial intelligence-based cache distribution, using machine learning algorithms across core, fog, and customer data centers to predict optimal data needs, synchronize data intelligently, and employ customized caching and push techniques based on customer behavior patterns, enabling proactive and reactive caching strategies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional machine learning methods are used for data caching decisions, then implementation is simpler, but they cannot account for heterogeneous products and usage patterns across diverse customer sites

Engineering Contradiction:

Improveability to account for heterogeneous products and usage patternsVSAvoidcomplexity of distributed machine learning approach

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the machine learning approach into multiple distributed components: local ML models at edge data centers, regional ML models at regional data centers, and centralized coordination at the core data center. Each segment handles specific aspects of cache management for its local context, enabling adaptability to heterogeneous products and usage patterns while distributing the computational complexity across multiple locations rather than concentrating it all at one place.

Inventive Principle:

Principle #1Segmentation

2Loss of information

If data is synchronized frequently across all customer data centers, then data freshness is improved, but network congestion increases

Engineering Contradiction:

Improvedata freshness and availabilityVSAvoidnetwork bandwidth consumption

Core Design Contradiction:

Loss of informationVSLoss of energy

Solution Approach 1:

The patent implements local quality by allowing each edge data center to maintain its own cached data with locally-adapted ML models that predict usage patterns specific to that location. This enables each site to have fresh, relevant data without requiring constant synchronization with all other sites. The regional data centers then synchronize only aggregated predictions and updates with the core data center, significantly reducing network bandwidth consumption while maintaining data freshness where needed.

Inventive Principle:

Principle #3Local quality

3Adaptability or versatility

If customized caching strategies are implemented at each customer site, then customer-specific requirements are met, but system complexity increases

Engineering Contradiction:

Improvecustomer-specific data management capabilityVSAvoidcomplexity of distributed cache management system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal ML framework that can be deployed across all customer sites, with the same basic architecture and algorithms adapted to local conditions through training on local data. This multi-functional system serves multiple purposes: it handles customer-specific customization needs while also performing aggregation and coordination functions at regional and core levels. The universal framework reduces complexity by providing a standardized approach that works across diverse customer requirements rather than requiring completely separate systems for each site.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Speed

If more data is cached at edge data centers, then response performance is improved, but storage requirements and data management complexity increase

Engineering Contradiction:

Improveresponse performanceVSAvoiddata storage volume

Core Design Contradiction:

SpeedVSQuantity of substance

Solution Approach 1:

The patent applies partial action by caching only the subset of data that is most likely to be needed at each edge data center, as determined by local ML predictions. Rather than caching all possible data, the system uses machine learning to identify and cache only the relevant portion needed for local operations. This reduces storage requirements while still improving response performance for the critical data subsets. The system performs excessive action in terms of prediction accuracy, using sophisticated ML models to ensure the cached data is highly relevant.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11657112B2Artificial intelligence-based cache distribution

Publication Date: 2023.05.23 DELL PROD LP

AI summary

Techniques are disclosed for data management techniques using artificial intelligence-based cache distribution within a distributed information processing system. For example, a cohesive and distributed machine learning approach between the same or similar customer data centers and products predict optimal data needed at each customer data center, and intelligently synchronize or federate the data between customer data centers and a core data center using a combination of customized caching and push techniques according to one or more customer behavior patterns.