Dynamic Resource Allocation via Multi-Period Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing resource scheduling technologies in cloud environments face inefficiencies due to over-estimation of resource needs by users, leading to significant resource waste, especially in large-scale systems with diverse workloads, and fail to optimize resource utilization across heterogeneous scenarios.
Innovation Solution
A resource allocation method that constructs long-period and short-period resource feature maps based on usage data to allocate resources effectively, ensuring accurate reflection of workload demands and improving resource utilization rates through unified processing and resource overselling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If request-based static resource model is used, then resource allocation is simple, but resource utilization rate deteriorates due to over-estimation of resource needs
Solution Approach 1:
The patent transitions from static resource allocation to dynamic resource allocation by introducing long-period and short-period resource feature maps that adapt to changing workload characteristics. The system continuously learns from historical resource usage data and adjusts allocation strategies based on actual workload patterns, enabling resources to be dynamically matched to demand rather than statically over-provisioned.
Solution Approach 2:
The patent changes the parameters of resource allocation by introducing time-period differentiation (long-period vs. short-period features) and workload-type classification. Instead of using a single static allocation parameter, the system uses multiple feature maps with different temporal characteristics to capture diverse workload patterns, enabling more precise resource matching.
2Ease of operation
If manual resource estimation is performed, then resource allocation can be controlled, but accuracy deteriorates due to difficulty in estimating actual resource needs
Solution Approach 1:
The patent implements feedback mechanisms by continuously monitoring actual resource usage and comparing it with allocated resources. The system uses historical resource usage data to train feature maps that predict future resource needs more accurately. This closed-loop feedback enables the system to learn from past estimation errors and improve prediction accuracy over time.
Solution Approach 2:
The system enables workloads to effectively self-service by automatically learning their own resource consumption patterns through the feature maps. Each workload type develops its own characteristic feature representation, allowing the system to automatically allocate appropriate resources without manual intervention or expert estimation.
3Ease of manufacture
If single-machine matching rule is used, then resource allocation is straightforward, but resource utilization deteriorates due to severe resource fragmentation
Solution Approach 1:
The patent introduces a universal resource pool that serves multiple workload types simultaneously, rather than dedicating resources to single machines or workloads. The long-period and short-period feature maps enable the system to manage diverse workload types within a unified allocation framework, allowing fragmented resources to be consolidated and allocated efficiently across different workload categories.
Solution Approach 2:
The patent segments resource allocation into different time periods (long-period and short-period) and workload types, allowing fragmented resources to be managed and allocated in distinct temporal and functional categories. This segmentation enables the system to handle resource fragmentation by organizing scattered resources into structured feature maps that can be efficiently matched to corresponding workload segments.
Data Source
AI summary
Provided in the present application are a resource allocation method and apparatus. The resource allocation method comprises: acquiring resource usage data corresponding to work loads; according to the resource usage data, constructing a long-period resource feature map corresponding to a long-period load type, and a short-period resource feature map corresponding to a short-period load type; according to the short-period resource feature map, allocating, from resources to be allocated, short-period resources for the short-period load type; and according to the long-period resource feature map, allocating, from the short-period resources, long-period resources for the long-period load type.