Periodicity-Aware Predictive Modeling for Distributed Storage Resource Allocation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Legacy techniques for managing resources in distributed virtualization systems fail to accurately account for seasonal or periodically recurring resource usage characteristics, leading to misallocation of resources due to reliance on fixed analysis windows that do not capture dynamic behavior.

Innovation Solution

Implementing periodicity-aware predictive modeling to determine resource allocation by dynamically identifying and using training windows that capture periodic patterns in historical resource usage data, allowing for more accurate prediction of resource demands and efficient allocation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If fixed analysis windows are used to determine resource allocation, then implementation simplicity is maintained, but resource allocation accuracy deteriorates due to inability to capture periodic patterns

Engineering Contradiction:

Improveresource allocation mechanismVSAvoidresource usage prediction accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent transforms the static fixed analysis window into a dynamic training window that adapts to periodic patterns in resource usage. The system automatically adjusts the training window duration based on detected periodicity, allowing the resource allocation mechanism to respond dynamically to changing usage patterns while maintaining implementation feasibility through automated detection algorithms.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent introduces periodicity-aware analysis that specifically looks for and adapts to recurring patterns in resource usage. By detecting the periodic nature of workload patterns and adjusting the training window accordingly, the system captures seasonal and recurring usage characteristics that fixed windows miss, thereby improving prediction accuracy without excessive complexity.

Inventive Principle:

Principle #19Periodic action

2Measurement precision

If longer fixed analysis windows are used to capture more historical data, then more periodic patterns may be captured, but resource allocation responsiveness to current demands deteriorates

Engineering Contradiction:

Improveperiodic pattern detection capabilityVSAvoidresource allocation responsiveness

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system dynamically determines the optimal training window duration based on detected periodicity characteristics rather than using a static long window. This allows the analysis period to adapt - extending when needed to capture periodic patterns and contracting when recent data is sufficient, thereby balancing pattern detection capability with responsiveness to current demands.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system performs preliminary periodicity detection and training window determination before resource allocation decisions are made. By pre-analyzing the periodic characteristics of resource usage and establishing an appropriate training window in advance, the system prepares accurate predictions without delaying the actual resource allocation response to current demands.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If periodicity-aware predictive modeling is implemented, then resource allocation accuracy improves, but computational complexity increases

Engineering Contradiction:

Improveresource demand prediction accuracyVSAvoidpredictive modeling system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system implements automated periodicity detection and training window determination that operates without manual intervention. The predictive modeling system automatically identifies periodic patterns, determines appropriate analysis durations, and adjusts resource allocation predictions accordingly, reducing the operational complexity burden despite the enhanced analytical capabilities.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system incorporates feedback mechanisms that continuously monitor resource usage patterns and adjust the training window and predictions based on observed periodicity. This self-adjusting feedback loop improves prediction accuracy over time while managing complexity through automated adaptation rather than requiring complex manual configuration and tuning.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10484301B1Dynamic resource distribution using periodicity-aware predictive modeling

Publication Date: 2019.11.19 NUTANIX INC

AI summary

Resource allocation techniques for distributed data storage. A set of distributed storage system historical resource usage measurements are collected and stored using distributed storage system measurement techniques. The resource usage metrics are associated with and/or derived from processing entities in the distributed storage computing system. An analysis module determines a training window time period corresponding to a portion of the collected distributed storage system historical resource usage measurements. The training window time period is determined so as to provide an earlier time boundary and a later time boundary that defines a periodically recurring portion of the distributed storage system historical resource usage measurements. A latest cycle of those periodically recurring measurements are then used to train a predictive model, which in turn is used to produce distributed storage system predicted resource usage characteristics. Resource allocation decisions are made based at least in part on predictions from the trained predictive model.