Workload Management via Dynamic Resource Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workload management systems struggle to effectively differentiate and manage batch and interactive workloads due to static identifiers not correlating well with process operating characteristics, and changes in workload characteristics over time, leading to resource interference and performance degradation.
Innovation Solution
Characterizing processes based on actual resource consumption history and classifying them into resource consumption classes to dynamically adjust resource allocation and assignment, allowing for segregation of interactive and batch workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If static identifiers (executable name, user account) are used to classify and assign processes to workloads, then the workload management system can automatically group and manage processes, but the classification accuracy deteriorates because static identifiers do not correlate well with process operating characteristics
Solution Approach 1:
The patent changes the classification parameters from static identifiers (executable name, user account) to dynamic parameters based on actual resource consumption history. The system monitors resource usage patterns and classifies processes according to their observed behavior rather than predetermined labels, thereby improving classification accuracy while maintaining automation.
Solution Approach 2:
The system allows processes to be automatically classified based on their own resource consumption patterns without requiring manual intervention or predefined static identifiers. Each process essentially classifies itself through its observed behavior, with the workload management software detecting and categorizing processes based on their actual resource usage characteristics.
2Device complexity
If processes are assigned to workloads based on static identifiers, then the management system can maintain simple workload definitions, but the system becomes inflexible when workload characteristics change over time
Solution Approach 1:
The patent introduces dynamic classification that automatically adapts to changing workload characteristics. Instead of fixed assignments based on static identifiers, the system continuously monitors resource consumption patterns and reclassifies processes as their behavior changes, enabling the management structure to remain simple while becoming highly adaptable to temporal variations in workload characteristics.
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring resource consumption history and using this information to dynamically adjust process classifications. This feedback loop allows the workload management system to automatically respond to changing workload characteristics without increasing structural complexity or requiring manual reconfiguration.
3Productivity
If batch and interactive workloads are managed together in the same operating system instance, then resource utilization can be maximized, but batch workloads starve interactive workloads of resources, degrading user responsiveness
Solution Approach 1:
The patent applies segmentation by classifying processes into different workload types (batch, interactive, transactional) based on their resource consumption patterns and managing them with different resource management parameters. This allows the system to maintain high overall resource utilization while ensuring that interactive workloads receive sufficient resources to maintain user responsiveness, as each segment can be optimized independently.
Data Source
AI summary
A method of allocating computer resources to workloads involves monitoring utilization of computer resources by workloads to generate utilization as a function of time data. The utilization data is used to classify workloads into demand types as a function of the utilization data. A workload manager plans and implements an allocation of computer resources to the workloads in part as a function of the demand types of the workloads.