Workload Resource Allocation via Probability Distributions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workload management systems only consider a single best prediction for resource allocation, failing to account for supplementary predictions that reflect additional knowledge about workload operations, leading to inefficiencies in resource distribution and potential under-allocation to workloads with uncertain resource needs.
Innovation Solution
Generating additional allocation requests based on multiple resource requirement predictions, each assigned a lower priority, to allocate resources according to the likelihood of occurrence, ensuring that resources are allocated to meet both certain and uncertain workload demands effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single best prediction is used for resource allocation, then the allocation process is simple and fast, but resource allocation does not account for supplementary predictions and may lead to under-allocation for workloads with uncertain resource needs
Solution Approach 1:
The patent segments the resource allocation process into multiple priority levels, where each level corresponds to a different prediction scenario (best prediction, supplementary predictions). This segmentation allows the system to handle multiple predictions systematically without overwhelming complexity, resolving the contradiction between allocation efficiency and processing complexity.
Solution Approach 2:
The patent performs preliminary action by pre-establishing multiple priority levels for resource requests based on different prediction outcomes. This preliminary structuring of allocation priorities allows the system to efficiently process multiple predictions without increasing runtime complexity, as the priority framework is prepared in advance.
2Reliability
If multiple resource requirement predictions are considered, then resource allocation better matches actual workload needs, but the allocation process becomes more complex
Solution Approach 1:
The patent segments the allocation process into distinct priority levels (first priority for best prediction, second priority for supplementary predictions). This segmentation maintains reliability by considering multiple predictions while managing complexity through structured prioritization, where each segment handles a specific prediction scenario independently.
Solution Approach 2:
The patent applies partial action by generating and processing multiple resource requirement predictions, including supplementary predictions that may not all be fulfilled. This partial/excessive approach ensures that resource allocation accounts for uncertain workload needs, improving reliability while the complexity is managed through selective fulfillment based on available resources.
3Speed
If resources are allocated based on single best prediction, then allocation speed is maintained, but resource shortages may occur for workloads with uncertain resource needs
Solution Approach 1:
The patent performs preliminary action by pre-calculating multiple resource requirement scenarios and establishing priority levels before actual allocation occurs. This preliminary preparation allows the system to maintain allocation speed while improving reliability, as the multi-prediction framework is ready to be applied without additional runtime computation delays.
Solution Approach 2:
The patent segments the allocation process into priority levels that can be processed sequentially. By segmenting the consideration of different predictions into structured levels, the system maintains efficient allocation speed while ensuring resource sufficiency through comprehensive consideration of multiple scenarios at appropriate priority levels.
Data Source
AI summary
A computer system allocates computer resources to workloads by generated plural requests per combination of workload and allocation period from probability distributions generated at least in part from utilization data.