Multivariate k-NN Forecasting for Network Resource Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing approaches to forecasting computer infrastructure needs for enterprises often lead to over-provisioning, resulting in underutilized servers and increased capital and operating expenses, as they focus on univariate time series without considering correlations among multiple related metrics.
Innovation Solution
The implementation of multivariate k-nearest neighbor (k-NN) forecasting and adaptive scheduling algorithms that group correlated metrics to generate multi-step ahead predictions, allowing for dynamic scaling of network resources based on historical and current load patterns, thereby reducing waste and optimizing resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If univariate time series forecasting is used, then the forecasting system is simple to implement, but the prediction accuracy is insufficient because correlations among multiple related metrics are ignored
Solution Approach 1:
The patent transitions from univariate to multivariate time series forecasting by incorporating multiple related metrics (CPU load, memory utilization, disk I/O, network traffic) simultaneously. This dimensional expansion allows the system to capture correlations among metrics, significantly improving prediction accuracy while managing complexity through structured approaches like variable grouping and feature selection
2Productivity
If multivariate k-NN forecasting with variant grouping is implemented, then prediction performance and speed are improved, but the computational complexity increases due to correlation analysis and grouping operations
Solution Approach 1:
The patent segments the multivariate time series data by grouping variables based on correlation analysis. This segmentation divides the complex multivariate forecasting problem into smaller, more manageable sub-problems that can be processed more efficiently by the k-NN algorithm, thereby improving prediction speed while reducing the overall computational burden
Solution Approach 2:
The patent performs preliminary correlation analysis and variable grouping before executing the k-NN forecasting algorithm. This pre-processing step organizes the data structure in advance, enabling faster computation during the actual forecasting phase by reducing the search space and eliminating redundant calculations
Data Source
AI summary
Computer-implemented systems and methods forecast network resource and/or infrastructure needs for an enterprise computer system that employs network servers to host resources that are requested by network users. Based on the forecasts, the network resources can be scaled or provisioned accordingly. The state of the networkservers can be dynamically adjusted to meet the request needs of the users while reducing excess capacity. The forecasting techniques are also applicable to cloud computing environments. Based on the forecasts, the cloud server pool can be scaled dynamically, so that the system's scale satisfies the changing requests and avoids wasting resources when the system is under low load.