Proximate Compute Offloading for Resource-Intensive AI Tasks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
AI models require extensive computational resources that are often not available on single computing devices, leading to resource constraints when executed locally.
Innovation Solution
A system and method for offloading resource-intensive AI tasks to proximate computing devices by detecting and selecting suitable devices based on weighted scores derived from device advertisement packets, considering factors like processing power, cache availability, and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI models are deployed locally on a single computing device, then cloud dependency is reduced and local processing capability is improved, but resource constraints prevent efficient execution of resource-intensive AI tasks
Solution Approach 1:
The system segments the AI compute task across multiple proximate devices instead of relying on a single device. Each device contributes its available resources (CPU, GPU, cache, storage) to collectively execute the AI model, thereby overcoming individual resource constraints while maintaining local processing capability
Solution Approach 2:
The system merges computational resources from multiple proximate devices to form a distributed compute cluster. By combining processing power, memory, and storage across devices, the system achieves sufficient resources to execute resource-intensive AI tasks efficiently without cloud dependency
2Reliability
If resource-intensive AI tasks are executed on a single local device, then data privacy and security are improved, but the device's computational resources are insufficient
Solution Approach 1:
The AI compute task is segmented and distributed across multiple local devices that are already in proximity and trusted by the user. This segmentation allows the system to aggregate computational power while keeping all processing local, thereby maintaining data privacy and security without sacrificing computational capability
Solution Approach 2:
Multiple devices are utilized for a single AI compute task, making each device's resources serve a universal purpose. The system can dynamically allocate tasks across available devices based on their current resource availability, thereby maximizing the use of existing local computational resources
3Power
If multiple proximate devices are utilized for AI compute tasks, then available computational resources are improved, but device selection and task distribution complexity increases
Solution Approach 1:
Each proximate device autonomously advertises its available computational resources (CPU capacity, GPU availability, cache size, storage) to the system. This self-service mechanism eliminates the need for complex centralized resource discovery and enables automatic task allocation based on real-time device capabilities
Solution Approach 2:
The system continuously monitors resource availability from proximate devices through advertisement packets and uses this feedback to dynamically allocate AI compute tasks. Devices provide real-time feedback on their capacity, and the system adjusts task distribution accordingly, simplifying the complexity of resource management through continuous information exchange
Data Source
AI summary
The technology disclosed herein provides a method of detecting proximate devices and extending a resource-intensive AI task to one of the proximate devices, the method including determining that a compute task on a primary device is a resource-intensive AI task requiring one or more resources above a threshold, determining that the resource-intensive AI task can be delegated to one or more proximate computing devices, scanning one or more proximate devices to receive device advertisement packets, determining weighted scores for the one or more proximate devices based on the advertisement packets, selecting one of the one or more proximate devices based on the weighted scores, and communicating a compute task delegation request to the selected proximate device.