Workload Placement Optimization Using Interaction Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large networks with multiple switches and tiers, the placement of computing workloads is often not optimal, leading to increased latency and bandwidth utilization, resulting in lower overall performance.
Innovation Solution
A method that determines attributes of a computing workload, predicts its interaction with other workloads within the network, and uses network topology to optimize placement, employing a workload manager to deploy virtual machines based on predicted interaction probabilities and network layout.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If computing workloads are placed in large networks with many switches and multiple tiers, then network coverage and connectivity are improved, but latency and bandwidth utilization increase, resulting in lower performance
Solution Approach 1:
The system performs preliminary analysis of workload attributes and network topology before placement, predicting interactions and determining optimal locations in advance. This prevents suboptimal placements that would cause latency issues, as the workload manager evaluates compatibility and network conditions before committing workloads to specific hosts.
Solution Approach 2:
The system places workloads in specific local network locations based on their interaction requirements. By analyzing workload attributes and predicting interactions, the system identifies optimal local placements that minimize latency for specific workload pairs, rather than using uniform placement strategies across the entire network.
2Area of stationary object
If computing workloads are placed in large networks with many switches and multiple tiers, then network coverage and connectivity are improved, but bandwidth utilization increases, resulting in lower performance
Solution Approach 1:
The system performs preliminary analysis of workload attributes and network topology before placement, predicting interactions and determining optimal locations in advance. This prevents suboptimal placements that would cause bandwidth congestion, as the workload manager evaluates compatibility and network conditions before committing workloads to specific hosts.
Solution Approach 2:
The system places workloads in specific local network locations based on their interaction requirements. By analyzing workload attributes and predicting interactions, the system identifies optimal local placements that minimize bandwidth consumption for specific workload pairs, rather than using uniform placement strategies across the entire network.
3Ease of manufacture
If traditional workload placement methods are used, then implementation simplicity is maintained, but network performance and efficiency deteriorate
Solution Approach 1:
The workload manager automatically analyzes workload attributes, predicts interactions, and determines optimal placements without manual intervention. The system self-configures by evaluating network topology and workload compatibility, eliminating the need for manual placement optimization while achieving improved network performance through automated intelligent decision-making.
Solution Approach 2:
The system uses feedback from workload attribute analysis and interaction predictions to continuously optimize placement decisions. By monitoring workload characteristics and network conditions, the workload manager adjusts placement strategies to maintain optimal performance, creating a closed-loop system that adapts to changing conditions.
Data Source
AI summary
Systems and methods for determining placement of computing workloads within a network are disclosed. According to an aspect, a method includes determining one or more attributes of a computing workload to be placed within a network. The method also includes predicting interaction of the computing workload with one or more other computing workloads within the network based on the one or more attributes. Further, the method includes determining placement of the computing workload within the network based on the predicted interaction and a topology of the network.