Fog Network Load Transfer Using Graph Morphism and Node Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fog networks face challenges in managing computation load when a fog node becomes unavailable due to factors like leaving the network, entering an inactive state, or losing connectivity, leading to potential service disruptions and inefficient reorganization of computation tasks.
Innovation Solution
A method involving Mutual Information-based node selection and a learning model to identify a morphism that transfers computation from an unavailable fog node to a candidate set of nodes, preserving the computation graph while minimizing loss, using morphing operations like length, width, and subnet morphing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computation load is transferred to other fog nodes when a node becomes unavailable, then service continuity is maintained, but the complexity of managing computation dependencies increases
Solution Approach 1:
The patent applies preliminary action by proactively predicting which fog nodes are likely to become unavailable before they actually fail. The system identifies candidate replacement nodes in advance and prepares computation transfer plans, so when a node actually becomes unavailable, the transfer can occur immediately without complex real-time decision-making. This resolves the contradiction by maintaining service continuity through pre-planned transfers while avoiding the complexity of managing dependencies during actual failure events.
Solution Approach 2:
The patent introduces a controller as an intermediary between fog nodes and the cloud server. This controller manages the computation dependency information and coordinates the transfer of computation tasks when nodes become unavailable. By centralizing the management of computation dependencies in this intermediary controller, the system maintains service continuity while reducing the complexity that would otherwise be distributed across all fog nodes.
2Reliability
If computation is shifted to maintain service continuity, then reliability improves, but the time required for reorganization increases
Solution Approach 1:
The system performs preliminary identification of candidate replacement nodes and pre-establishes computation dependency information before failures occur. When a fog node becomes unavailable, the transfer can immediately proceed to a pre-identified candidate node using pre-analyzed dependency information, dramatically reducing the reorganization time while maintaining service continuity.
Solution Approach 2:
The patent implements beforehand cushioning by maintaining a pool of candidate replacement nodes that are pre-evaluated and ready to accept computation tasks. This cushioning layer of pre-prepared resources ensures that when failures occur, the system can immediately transfer computation without time-consuming searches or evaluations, thus maintaining reliability while minimizing reorganization time.
3Device complexity
If fog nodes work independently, then device complexity is reduced, but the ability to predict churn and manage load transfer decreases
Solution Approach 1:
The controller serves as an intermediary that centralizes the complex tasks of monitoring fog node status, predicting churn, and managing computation dependencies. Individual fog nodes can continue to work independently with simple operations, while the intermediary controller handles the sophisticated churn prediction and coordination, thus maintaining node simplicity while enabling advanced adaptability.
Solution Approach 2:
The controller is designed as a universal component that performs multiple functions: monitoring fog node status, predicting churn, managing computation dependencies, and coordinating load transfers. By consolidating these diverse functions in a single multi-functional controller, the system maintains simple independent fog nodes while achieving sophisticated overall adaptability and churn prediction capability.
Data Source
Figure 1
Figure 2a
Figure 2b
AI summary
A method (100) for managing computation load of a fog node belonging to a fog network is disclosed, a computation capacity of the fog node being predicted to become unavailable to the fog network. The method comprises identifying, from among nodes of the fog network, a candidate set of nodes for computational load transfer from the fog node, wherein the candidate set is identified on the basis of Mutual Information between the fog node and other nodes of the fog network (110). The method further comprises obtaining a computation graph representing computation in the fog network (120), and using a learning model to identify a morphism from the obtained computation graph to a new computation graph, in which the fog node is not included, which morphism preserves computation of the obtained computation graph while minimising a loss function (130). The identified morphism comprises a sequence of one or more morphing operations that replaces the fog node in the obtained computation graph with a topology of one or more nodes selected from the candidate set (130a). The method further comprises causing computation performed at the fog node to be transferred to one or more nodes of the candidate set in accordance with the identified morphism (140).