Cloud Scheduler Multi-Resource Allocation with Ellipsoidal Uncertainty
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional multi-resource allocation policies in cloud computing fail to account for dynamic changes in resource demands, leading to inefficient and unfair resource allocation, resulting in performance losses and system robustness issues.
Innovation Solution
A multi-resource scheduling method that employs an ellipsoidal uncertainty model and non-linear optimization techniques, using fairness and efficiency formulas to dynamically adjust resource allocation based on changing demands, ensuring optimal fairness and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional resource allocation policies allocate resources based on fixed bundles or dominant resource fairness, then resource allocation simplicity is maintained, but adaptability to dynamic demand changes deteriorates
Solution Approach 1:
The patent implements dynamic resource allocation by transitioning from static bundle-based allocation to continuous adjustment based on real-time demand. The allocation policy dynamically modifies resource distribution across cloud users as demand patterns change, ensuring adaptability without requiring complete policy redesign. This resolves the contradiction by making the system responsive to dynamic conditions while maintaining manageable complexity through incremental adjustments.
Solution Approach 2:
The patent changes key allocation parameters from fixed bundle sizes to flexible, demand-responsive quantities. By allowing allocation parameters to vary continuously based on observed demand patterns rather than remaining static, the system achieves adaptability to dynamic conditions. This parameter flexibility resolves the contradiction between adaptability and complexity by using simple parameter adjustments rather than complex policy transformations.
2Reliability
If resource allocation estimates exceed actual demand to ensure timely job completion, then service reliability is improved, but resource utilization efficiency deteriorates
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor actual resource consumption and allocation effectiveness. By comparing estimated allocations with actual demand patterns, the system adjusts future allocations to achieve tighter fits between supply and demand. This feedback loop resolves the contradiction by reducing over-allocation while maintaining service reliability through data-driven adjustments rather than conservative over-provisioning.
Solution Approach 2:
The patent uses historical demand data and patterns to make preliminary allocation decisions that anticipate future needs without excessive over-provisioning. By analyzing past consumption patterns, the system pre-positions resources more accurately, reducing the need for conservative over-allocation. This preliminary action based on empirical patterns resolves the contradiction by achieving reliability through informed prediction rather than excessive resource reservation.
3Productivity
If resource allocation maximizes resource utilization to improve economic profit, then productivity is improved, but fairness among cloud users deteriorates
Solution Approach 1:
The patent applies different allocation strategies to different cloud users based on their specific characteristics, demand patterns, and service level agreements. Rather than uniform allocation, the system tailors resource distribution to local conditions of each user while maintaining overall system efficiency. This local differentiation resolves the contradiction by allowing aggressive allocation to efficient users while providing protected allocation to users requiring fairness guarantees.
Solution Approach 2:
The patent dynamically adjusts allocation parameters for different users based on their utilization efficiency and fairness metrics. By changing allocation weights, priorities, and constraints based on observed performance and fairness requirements, the system balances overall productivity with individual user fairness. This parameter flexibility resolves the contradiction by allowing the system to optimize for productivity when appropriate while maintaining fairness when required.
Data Source
AI summary
The present invention provides a multi-resource scheduling method responding to uncertain demands in a cloud scheduler, where two computation formulas for fairness and efficiency are used as cost functions in an optimization problem. For some change sets with uncertain resource demands, a robust counterpart of an original non-linear optimization problem is computationally tractable. Therefore, the present invention models features of these sets with uncertain resource demands, i.e., establishes an ellipsoidal uncertainty model. In this model, each coefficient vector is put into a hyper-ellipsoidal space and used as a metric to measure an uncertainty degree. With the ellipsoidal uncertainty model, a non-linear optimization problem is solved and a resource allocation solution that can respond to dynamically changing demands can be obtained.