Scale Out Workload Management via Dynamic Node Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current workload management in distributed computing systems relies on manual methods, which are slow and inefficient, making it difficult for systems to adapt quickly to changing resource demands, resulting in suboptimal resource allocation and increased energy consumption.
Innovation Solution
A scale out workload management system that automatically monitors and adjusts resource allocation across multiple nodes by using an application manager and a workload manager to collect, aggregate, and compare performance metrics against targets, enabling dynamic startup and shutdown of nodes to optimize resource usage and minimize energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual methods are used for workload assignment, then operational simplicity is maintained, but resource allocation efficiency deteriorates
Solution Approach 1:
The workload management system performs self-service by automatically monitoring performance metrics, comparing them against targets, and dynamically allocating resources without human intervention. The system monitors its own state and makes autonomous decisions to optimize resource distribution across nodes based on real-time workload demands.
Solution Approach 2:
The patent replaces manual mechanical workload assignment with an automated electronic monitoring and control system. The mechanical process of manual resource allocation is substituted with electronic performance monitoring, automated comparison logic, and dynamic resource provisioning mechanisms that respond automatically to system conditions.
2Device complexity
If manual workload management is used, then system complexity is reduced, but adaptability to changing demands deteriorates
Solution Approach 1:
The system implements continuous feedback loops by monitoring performance information from nodes, comparing actual performance against target performance levels, and dynamically adjusting resource allocation based on the feedback. This closed-loop control enables the system to automatically adapt to changing workload demands while maintaining manageable complexity through standardized feedback mechanisms.
Solution Approach 2:
The workload management system transitions from static manual allocation to dynamic automated allocation. Resources are dynamically provisioned and de-provisioned based on real-time performance monitoring and comparison against targets, allowing the system to adapt flexibly to changing demands without requiring complex manual reconfiguration.
3Extent of automation
If manual resource assignment is used, then automation overhead is minimized, but response speed to demand changes deteriorates
Solution Approach 1:
The system maintains continuous monitoring of performance information and continuous comparison against targets, ensuring uninterrupted detection of demand changes. This continuous useful action eliminates the delays inherent in manual periodic assessments, enabling rapid response to workload fluctuations while keeping automation overhead manageable through efficient continuous monitoring protocols.
Data Source
AI summary
A scale out workload management system executing on a computer system controls simultaneous execution of multiple workloads. The computer system includes multiple nodes, where each node defines computer resources consumable by a workload. The scale out workload management system includes an application manager coupled to the nodes. The application manager controls start up and shut down of the nodes based on demands placed on the workloads. The scale out workload management system further includes a workload manager coupled to the application manager. The workload manager includes a monitor module that collects performance information from nodes on which the workloads are executing, an aggregator module that collects the performance information and aggregates the performance information from all nodes supporting the workloads, and a comparator module that compares aggregated performance information for the workloads to a performance target and determines if an allocation of nodes to the workloads should be changed.