Data Center Workload Scheduling Using Localized Weather Predictions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data center management systems face inefficiencies and increased costs due to inaccurate weather predictions for large geographical areas, leading to disruptions and financial penalties from incorrect power scheduling.

Innovation Solution

A computer-implemented method using a data center management system to predict highly localized weather conditions, determine potential degradations, and schedule computational workloads based on priorities to optimize resource usage and prevent disruptions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If weather conditions are predicted for a large geographical area, then the prediction coverage is sufficient, but the prediction accuracy and specificity are insufficient

Engineering Contradiction:

Improvegeographical coverage areaVSAvoidweather prediction accuracy

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent divides the large geographical area into multiple smaller grid cells or zones, each with its own weather prediction model. This segmentation allows the system to maintain broad coverage while providing location-specific accuracy for each smaller region, directly resolving the contradiction between coverage area and prediction precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements local quality by tailoring weather predictions to specific locations within the geographical area. Each location receives customized prediction data based on its unique characteristics, ensuring high accuracy for power generation forecasting while maintaining overall regional coverage through the distributed prediction framework.

Inventive Principle:

Principle #3Local quality

2Reliability

If computational workloads are rescheduled due to incorrect power predictions, then operational disruptions are addressed, but computing resources and costs are consumed

Engineering Contradiction:

Improveoperational continuityVSAvoidcomputing resource consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent applies preliminary action by using accurate localized weather predictions to proactively schedule computational workloads before power supply issues occur. The system forecasts power availability in advance and pre-schedules workloads to match predicted power supply, preventing disruptions without needing to reschedule, thus maintaining reliability while avoiding the energy cost of reactive rescheduling operations.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback by continuously monitoring actual power generation against predictions and using this information to refine future workload scheduling decisions. This closed-loop system improves prediction accuracy over time, enabling more reliable operational planning and reducing the frequency and cost of corrective rescheduling actions.

Inventive Principle:

Principle #23Feedback

3Adaptability or versatility

If service level agreements are not met due to power prediction errors, then operational flexibility is maintained, but financial penalties are incurred

Engineering Contradiction:

Improveoperational flexibilityVSAvoidservice level agreement compliance

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent uses preliminary action by predicting power availability in advance and scheduling workloads to align with predicted supply before commitments are made. This proactive approach ensures service level agreements are met by design rather than by corrective action, maintaining both reliability and operational flexibility through advance planning based on accurate localized forecasts.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20240028395A1Managing an operation of a data center based on predicting localized weather conditions

Publication Date: 2024.01.25 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A data center management system may predict localized weather conditions for a location of a data center during a period of time. The data center management system may determine whether the localized weather conditions will degrade an operation of the data center during the period of time. The data center management system may obtain computational workload information identifying one or more computational workloads scheduled for execution at the data center during the period of time. The data center management system may determine, based on the computational workload information, one or more priorities associated with the one or more computational workloads. The data center management system may schedule the one or more computational workloads for execution based on the one or more priorities and based on determining whether the localized weather conditions will degrade the operation of the data center.