Workload Placement via Subgraph Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is challenging to determine the optimal portion of a device ecosystem for deploying a new workload instance due to the complexity of workload and infrastructure resource usage patterns.
Innovation Solution
A method for workload placement based on subgraph similarity, which involves encoding workload graphs and infrastructure subgraphs, performing subgraph similarity queries, and using machine learning predictions to determine the most suitable infrastructure for workload deployment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional workload placement methods are used, then deployment simplicity is maintained, but placement accuracy and resource optimization deteriorate
Solution Approach 1:
The patent introduces graph embeddings as an intermediary representation that captures complex workload-infrastructure relationships. The embedding model translates complex graph data into compact vectors that can be efficiently compared and ranked, serving as a mediator between the complex underlying system and the placement decision process.
Solution Approach 2:
The patent transforms the complex graph-based workload representation into numerical parameter space through graph embeddings. By converting graph structures into vector parameters, the system can efficiently perform similarity computations and rankings without directly manipulating complex graph data, thus improving accuracy while managing complexity.
2Productivity
If comprehensive infrastructure analysis is performed, then placement optimization improves, but processing time and computational resources increase
Solution Approach 1:
The patent pre-computes graph embeddings for infrastructure components and stores them in advance. When a workload needs placement, the system retrieves pre-computed embeddings and performs rapid similarity comparisons without re-analyzing the entire infrastructure graph, significantly reducing placement computation time while maintaining optimization quality.
Solution Approach 2:
The patent creates simplified numerical copies (embeddings) of complex graph structures that capture essential characteristics. These embedded representations serve as efficient proxies for the original complex data, enabling fast similarity computations and rankings without processing the full graph data during placement decisions.
3Reliability
If graph-based analysis is used, then workload placement accuracy improves, but system complexity and implementation difficulty increase
Solution Approach 1:
The patent replaces complex graph processing mechanisms with embedding-based vector comparisons. Instead of directly manipulating and comparing complex graph structures, the system substitutes graph analysis with numerical vector operations, which are computationally simpler and easier to implement while maintaining the ability to capture complex relationships.
Data Source
AI summary
Techniques described herein relate to systems and methods for workload placement based on subgraph similarity. Such techniques may include obtaining an encoded workload graph based on receiving a workload execution request; using the encoded workload subgraph to obtain encoded graphs representing previous workload executions, encoded subgraphs representing infrastructures on which the workload were executed, resource usage information, and execution metrics; using the encoded infrastructure subgraphs using subgraph similarity to identify candidate infrastructure subgraphs, using an ML model to predict an execution metric for an execution of the workload using the candidate; and selecting a best candidate infrastructure on which to execute the workload based on the predicted execution results.