Building Workload Orchestration for Secure Low-Carbon Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building management systems face challenges in efficiently managing computing workloads, optimizing machine learning model performance, and enhancing sustainability, particularly in the context of on-premises and off-premises devices, without considering environmental impacts.
Innovation Solution
A method for determining and assigning computing workloads to on-premises and off-premises devices based on workload characteristics and device capabilities, including edge devices, while considering sustainability impacts, such as carbon emissions, through workload partitioning and timing adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computing workloads are processed using on-premises building devices, then data security and control are improved, but computing resource capacity and flexibility deteriorate
Solution Approach 1:
The patent segments computing workloads into different categories (real-time control, data processing, analytics) and assigns them to appropriate devices (edge devices, cloud devices) based on their specific requirements. This allows the system to maintain data security for critical functions while leveraging cloud capacity for non-critical tasks.
Solution Approach 2:
The patent introduces a workload management system that acts as an intermediary between on-premises building devices and off-premises computing resources. This mediator intelligently routes workloads to appropriate devices, balancing security requirements with resource availability and flexibility.
2Adaptability or versatility
If computing workloads are processed using off-premises systems, then computing resource capacity and flexibility are improved, but data security and control deteriorate
Solution Approach 1:
The patent segments computing workloads into different categories and assigns them to appropriate devices (edge devices, cloud devices) based on their specific requirements. This allows the system to maintain data security for critical functions while leveraging cloud capacity for non-critical tasks.
Solution Approach 2:
The patent applies different quality requirements to different workloads based on their nature. Critical control functions receive local processing with high security requirements, while non-critical data processing can be offloaded to cloud devices with lower security requirements.
3Measurement precision
If machine learning models are continuously retrained to improve performance, then model accuracy is improved, but computing energy consumption and time increase
Solution Approach 1:
The patent implements periodic retraining of machine learning models based on performance degradation thresholds and time intervals, rather than continuous retraining. This allows the system to maintain adequate model accuracy while significantly reducing energy consumption associated with frequent retraining operations.
Solution Approach 2:
The patent changes the parameters of retraining operations by adjusting retraining frequency, data sampling rates, and model update intervals based on building operational patterns and energy availability, optimizing the balance between model accuracy and energy consumption.
4Speed
If computing workloads are processed during peak building operations, then operational responsiveness is improved, but carbon emissions increase
Solution Approach 1:
The patent schedules non-critical computing workloads to execute during periods of lower carbon intensity, while maintaining real-time processing for critical building operations. This periodic scheduling approach reduces overall carbon emissions while preserving operational responsiveness for essential functions.
Solution Approach 2:
The patent performs preliminary analysis of building operational patterns and carbon emission profiles to identify optimal processing windows for different workloads, allowing the system to proactively schedule tasks during low-emission periods without compromising operational requirements.
Data Source
AI summary
A method for processing compute activities of a building management system of a building, the method including receiving, by one or more processors, a selection from a user of a sustainability tuning level from among a plurality of sustainability tuning levels, the plurality of sustainability tuning levels representing different levels of weighting to be placed on mitigating a sustainability impact of computing workloads balanced against one or more other factors, receiving, by one or more processors, a computing workload to be processed by the building system, determining, by the one or more processors, an execution plan for processing the computing workload based at least in part on the selected sustainability tuning level, and causing, by the one or more processors, the computing workload to be processed in accordance with the execution plan.