Cloud Edge ML Operation Relocation via Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information processing systems face challenges in efficiently managing operations between cloud and edge platforms, particularly in determining when to relocate machine learning operations from cloud to edge platforms, leading to issues like high network bandwidth consumption, data privacy concerns, and delayed responses in critical situations.
Innovation Solution
The implementation of a method that uses machine learning algorithms to predict whether operations should be reallocated to the edge platform by determining if models are sufficiently trained and analyzing operational data such as data processing amounts and request frequencies, with the aid of serviceability prediction engines and orchestration engines that utilize smart contracts and blockchain technology for data sharing and resource management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If data processing operations are performed on a centralized cloud platform, then processing power and storage capacity are improved, but network bandwidth consumption increases and latency increases
Solution Approach 1:
The patent segments data processing operations by identifying specific operations (e.g., machine learning inference, data transformation) that can be separated from the centralized cloud platform and executed at edge locations. This segmentation allows processing power to be distributed, reducing network bandwidth consumption while maintaining cloud-based coordination and model management.
Solution Approach 2:
The patent implements local quality by deploying edge computing nodes at distributed locations closer to data sources. These edge nodes perform processing operations locally, reducing the need to transmit raw data over the network to the cloud, thereby decreasing network bandwidth consumption while maintaining processing capabilities.
2Power
If data processing operations are performed on a centralized cloud platform, then processing power is improved, but response time deteriorates
Solution Approach 1:
The patent segments processing operations into cloud-based tasks (model training, coordination) and edge-based tasks (inference, real-time processing). This segmentation enables time-sensitive operations to be executed locally at edge nodes, significantly reducing response time while the cloud platform maintains overall system coordination.
Solution Approach 2:
The patent implements local quality by placing edge computing resources geographically closer to data sources and end users. This local deployment reduces transmission latency and enables faster processing of time-critical operations while maintaining access to cloud-based processing power when needed.
3Power
If data is transmitted to cloud platform for processing, then processing capability is improved, but data privacy and security are compromised
Solution Approach 1:
The patent segments data handling by keeping sensitive raw data localized at edge nodes while only transmitting processed results or anonymized information to the cloud. This segmentation maintains data privacy and security by minimizing data transmission while still utilizing cloud processing capability for non-sensitive operations.
Solution Approach 2:
The patent implements local quality by maintaining data processing capabilities at edge locations, allowing organizations to keep sensitive data within their own infrastructure or controlled environments. This local processing preserves data privacy and security while still benefiting from cloud-based model updates and coordination when appropriate.
4Loss of energy
If machine learning operations are moved to edge platform, then network bandwidth consumption is reduced and response time is improved, but model training capability is reduced
Solution Approach 1:
The patent segments machine learning operations by keeping model training and retraining tasks on the cloud platform while deploying model inference and execution to edge nodes. This segmentation allows edge operations to reduce network bandwidth consumption and improve response time, while the cloud platform maintains full model training capability.
Solution Approach 2:
The patent implements local quality by deploying trained models to edge locations for inference operations, where they can process data locally without consuming network bandwidth. The cloud platform retains model training capability, creating a differentiated architecture where each location performs its specialized function optimally.
Data Source
AI summary
Techniques are disclosed for moving operations between cloud and edge platforms. For example, a method comprises executing a machine learning algorithm on a cloud platform and analyzing results of executing the machine learning algorithm. Based at least in part on the analysis, a determination is made whether the machine learning algorithm should be additionally trained. Based at least in part on a negative determination further execution of the machine learning algorithm is transferred from the cloud platform to an edge platform.