Classifier Chain for Shared Computing Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Infrastructure providers in shared computing environments face challenges in efficiently allocating resources to meet service level agreements (SLAs) without incurring excessive costs, as they need to balance resource allocation to avoid both over-provisioning and under-provisioning, which can lead to increased expenses or SLA violations.
Innovation Solution
A method using a trained machine learning classifier chain to predict the optimal combination of hardware types needed to satisfy SLA requirements by leveraging discriminative features from past applications and their corresponding resource allocations and SLA metrics, employing a classifier chain model that accounts for correlations between different hardware resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If more resources are allocated to applications, then SLA requirements are more likely to be satisfied, but management costs increase
Solution Approach 1:
The system changes the parameter of resource allocation from static/conservative to dynamic/optimal by using machine learning classifiers that analyze application features and predict the minimum resources needed to satisfy SLAs. This avoids both over-provisioning and under-provisioning, resolving the contradiction between reliability and cost.
Solution Approach 2:
The system enables self-service by automatically analyzing application characteristics, predicting resource requirements, and allocating resources without manual intervention. The classifier chain model autonomously makes allocation decisions that satisfy SLAs while minimizing costs, eliminating the need for conservative manual provisioning.
2Loss of energy
If fewer resources are allocated to minimize costs, then management costs decrease, but SLA requirements may be violated
Solution Approach 1:
The system transforms resource allocation from a conservative fixed parameter to a dynamically optimized parameter using machine learning. The classifiers predict the precise resource level needed for each application, ensuring SLA compliance while minimizing costs, thus resolving the contradiction between cost reduction and reliability maintenance.
Solution Approach 2:
The system implements feedback by using historical execution data and SLA measurements to continuously improve resource allocation predictions. The classifier chain learns from past allocations and outcomes, adjusting predictions to maintain SLA compliance while optimizing costs, resolving the contradiction between minimizing costs and ensuring reliability.
3Reliability
If conservative resource allocation policies are used, then SLA violations are avoided, but resource utilization efficiency decreases leading to increased costs
Solution Approach 1:
The system changes the allocation parameter from conservative estimates to ML-optimized predictions based on actual application behavior and features. This enables precise resource matching that maintains SLA compliance while maximizing utilization efficiency, resolving the contradiction between reliability and productivity.
Solution Approach 2:
The system performs preliminary analysis of application features and historical data before resource allocation to predict the optimal resource level. This preliminary action enables accurate upfront allocation that satisfies SLAs without requiring conservative over-provisioning, thus improving resource utilization efficiency while maintaining reliability.
4Device complexity
If non-linear relationships between resource combinations and SLA performance are not considered, then allocation complexity is reduced, but allocation accuracy decreases
Solution Approach 1:
The system replaces simple linear allocation rules with machine learning classifiers that automatically capture and model non-linear relationships between resource combinations and SLA performance. The ML models learn complex patterns from historical data, achieving high allocation accuracy without requiring manual specification of complex relationships, thus resolving the contradiction between complexity and precision.
Data Source
AI summary
Techniques are provided for allocation of shared computing resources using a classifier chain. An exemplary method comprises obtaining an application for execution in a shared computing environment having multiple resources with multiple combinations of one or more hardware types; obtaining discriminative features for the application; obtaining a trained machine learning classifier chain, wherein the trained machine learning classifier chain comprises multiple classifiers, wherein the multiple classifiers comprise a classifier for each combination of hardware types; and generating, using the at least one trained machine learning classifier chain, a prediction of the combination of hardware types needed to satisfy one or more service level agreement requirements for the application to be executed in the shared computing environment.