Resource Allocation Optimizing System for Cloud Workload Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing resource allocation methods in cloud infrastructure often result in surplus resource allocation to maintain service levels, leading to underutilization, as they cannot accurately predict workload changes or failures, necessitating twice the resources needed in a steady state.
Innovation Solution
A resource allocation optimizing system that performs predictive judgments based on time-series measured and predicted workloads and service levels to accurately allocate resources, using models like autoregressive integrated moving average (ARIMA) and neural networks to anticipate workload and response time, thereby optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If surplus resources are allocated to each service to maintain service level, then service reliability is improved, but resource utilization deteriorates
Solution Approach 1:
The system dynamically adjusts resource allocation based on predicted workload changes. Instead of static surplus allocation, resources are allocated adaptively according to time-series predictions of service workload, allowing the system to maintain service levels during peak demand while reducing allocation during low-demand periods.
Solution Approach 2:
The system performs preliminary predictions of workload changes using time-series analysis before actual demand occurs. By anticipating future workload based on historical patterns, the system can proactively allocate resources in advance, avoiding both surplus allocation and service level violations.
2Measurement precision
If accurate prediction of workload is implemented, then resource allocation precision is improved, but system complexity increases
Solution Approach 1:
The system introduces a prediction intermediary layer that sits between workload monitoring and resource allocation. This intermediary uses time-series analysis of historical workload data to generate predictions, which then inform resource allocation decisions. This mediator simplifies the overall system by decoupling direct monitoring from allocation while improving accuracy through pattern recognition.
Data Source
AI summary
It is necessary to allocate sufficient surplus resources to each service in order to maintain a service level of each service, but, as a result, many resources which are not used may be necessary.A resource allocation optimizing system performs, with respect to each of a plurality of services, a predictive judgement which is a determination of whether or not accurate prediction of a workload of the service is able to be anticipated, and controls allocation of a resource to the service on the basis of a result of the predictive judgement. With respect to each service, the predictive judgement is performed on the basis of at least one of (x) time-series measured workloads and time-series predicted workloads in a predefined period for the service and (y) at least one of a measured service level and a predicted service level in the predefined period for the service.