Cloud Management System for Predictive Workload Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud resources often experience temporary underutilization, leading to idle capacities, which can be leveraged to support additional workloads, but existing systems lack efficient methods to identify and aggregate these marginal resources for reuse.
Innovation Solution
A cloud management system that predicts workload demands and identifies underutilized resources by analyzing usage patterns, generating predictive marginal capacities, and allocating these resources to new or existing workloads through a subscription-based model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cloud resources are allocated to host workloads, then workload hosting capability is improved, but resource utilization efficiency deteriorates due to temporary underutilization and idle capacities
Solution Approach 1:
The patent merges underutilized resources from multiple host clouds into aggregated marginal capacities that can be shared and allocated to new workloads. The cloud management system combines idle resources across different clouds to create a pooled resource pool, enabling efficient reuse and improving overall utilization while maintaining workload hosting capabilities.
Solution Approach 2:
The system performs preliminary identification and aggregation of marginal capacities before they are needed. By proactively capturing and storing information about underutilized resources in advance, the system can quickly allocate these pre-identified capacities to new workloads when opportunities arise, eliminating delays and improving resource efficiency.
2Adaptability or versatility
If cloud resources are reserved for future workloads, then deployment flexibility is improved, but resource availability for current workloads deteriorates
Solution Approach 1:
The patent implements dynamic resource allocation where marginal capacities are identified and made available based on real-time workload demands. Rather than statically reserving resources, the system continuously monitors and adapts resource availability, allowing resources to be dynamically allocated to new workloads when host clouds have excess capacity without impacting current workload performance.
Solution Approach 2:
The system creates a universal pool of marginal capacities that can serve multiple purposes and different workload types. The aggregated resources from various host clouds can be allocated to diverse new workloads based on demand, providing universal access to underutilized resources across the entire cloud network rather than limiting them to specific predefined uses.
3Productivity
If marginal capacities are captured and aggregated across multiple clouds, then resource reuse efficiency is improved, but system complexity deteriorates
Solution Approach 1:
The patent introduces a cloud management system as an intermediary that handles the complexity of identifying, capturing, and aggregating marginal capacities across multiple clouds. This intermediary layer abstracts the complex multi-cloud coordination tasks, managing resource tracking and allocation while presenting a simplified interface to both host clouds and new workload requesters, thereby enabling efficient resource reuse without exposing the underlying system complexity.
4Measurement precision
If predictive workload estimation is implemented, then resource allocation accuracy is improved, but computational overhead deteriorates
Solution Approach 1:
The system applies partial predictive analysis by focusing computational efforts on identifying marginal capacities and estimating near-term workload demands rather than performing exhaustive long-term predictions. This selective approach achieves sufficient allocation accuracy for practical purposes while limiting computational overhead to only the necessary predictive calculations needed for effective resource capture and allocation.
Data Source
AI summary
Embodiments relate to systems and methods for generating cloud deployment targets based on predictive workload estimation. In aspects, a set of usage histories can store records for user workloads in a host cloud-based network recording the consumption of processor, memory, storage, operating system, application, or other resources subscribed to by the user. The operator of the cloud management system hosting the workloads of one or more users can track, identify, and manage the predictive marginal resource capacities of the set of host clouds, based on those historical usage patterns. The collective usage history can indicate, for instance, that a number of operating workloads tend to display a small under-utilization of processor or memory resources during certain overnight periods on a regular basis. The operator can then harvest those predictive marginal capacities, and offer a new user or workload a potential hosting subscription based on those expected resource availabilities.