Edge Tension Scores for Computing Infrastructure Workload Placement
Solution Overview
Problem
Existing computing infrastructure optimization techniques fail to consider interactions between neighboring nodes, leading to suboptimal workload placement and resource utilization in cloud computing environments.
Innovation Solution
The system generates edge tension scores for links between computing infrastructure nodes, taking into account the impact of neighboring nodes, to optimize initial workload placements, rebalancing, and reduce costs by identifying nodes with minimal tension for additional workloads and those with high tension for potential performance optimization through rebalancing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the orchestrator places workloads onto target compute hosts based on nominal capacity using bin packing scheduler, then the compute resource utilization is improved, but the overall system performance deteriorates due to ignoring network interfaces, memory architecture, and other related resources
Solution Approach 1:
The patent segments the resource evaluation into multiple independent components: compute capacity, network interface capacity, memory architecture characteristics, and storage I/O capacity. Each component is evaluated separately using specific metrics, and then combined to form a comprehensive suitability score for each compute host, allowing optimization of both compute utilization and overall system performance
Solution Approach 2:
The patent introduces multiple parameters beyond nominal compute capacity, including network interface bandwidth, memory bandwidth, storage I/O throughput, and latency metrics. These parameters are measured and used to dynamically adjust the workload placement decisions, transforming the single-parameter bin packing approach into a multi-parameter optimization system
2Ease of operation
If the orchestrator considers only compute resource for workload placement, then the selection process is simplified, but the resource utilization efficiency deteriorates due to ignoring related resources such as network interfaces and memory architecture
Solution Approach 1:
The patent performs preliminary characterization of each compute host's related resources (network interfaces, memory architecture, storage I/O) and stores this information in advance. When workload placement is needed, the orchestrator retrieves pre-computed metrics and suitability scores, avoiding complex real-time analysis while maintaining high resource utilization efficiency
Solution Approach 2:
The patent creates a virtual representation or model of the compute infrastructure that includes all relevant resources and their characteristics. This model is used for workload placement decisions, allowing the orchestrator to evaluate multiple factors without directly interacting with the physical resources during the selection process
3Reliability
If the system generates and analyzes edge tension scores for links between computing infrastructure nodes, then the workload placement optimization is improved by considering node interactions, but the computational complexity and processing time increase
Solution Approach 1:
The patent calculates edge tension scores selectively rather than for all possible node pairs. It focuses computation on edges that are likely to be relevant based on current workload patterns, network topology characteristics, and predefined thresholds, reducing the overall computational burden while maintaining placement optimization quality
Solution Approach 2:
The system uses feedback from monitoring actual workload performance and network conditions to adjust edge tension score calculations. Frequently updated metrics and changing workload patterns trigger selective recalculation of tension scores, rather than continuous full-system analysis, reducing processing time while maintaining optimization accuracy
Data Source
AI summary
A system comprises a scoring engine comprising at least one processor and memory. The scoring engine is to generate, based on telemetry information obtained from a plurality of nodes of a computing infrastructure, a first availability score for a first node of the plurality of computing infrastructure nodes and a second availability score for a second node of the plurality of computing infrastructure nodes. The scoring engine is further to generate, based on the first availability score of the first computing infrastructure node and the second availability score of the second computing infrastructure node, an edge tension score for a link between the first node and the second node.