IoT ML Container Image Distribution for Faster Node Startup
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning container image distribution solutions for Internet-of-Things devices face inefficiencies due to large image sizes and network bandwidth competition, leading to delayed container startup and reduced task execution efficiency.
Innovation Solution
A decentralized system with a master node and computing nodes, utilizing a machine learning model converter to convert models to ONNX format, a distributed storage module using the Bitswap protocol, and a scheduler to optimize image distribution across interconnected nodes, selecting the highest transmission rate for downloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the whole machine learning container image is downloaded, then the container can be started and tasks can be executed, but the large image size delays the starting time of the container and affects execution efficiency
Solution Approach 1:
The patent extracts only the essential machine learning model from the complete container image, separating it from the full computing framework. This allows the IoT device to download and execute the model without requiring the entire container image, significantly reducing download time while maintaining execution capability.
Solution Approach 2:
The machine learning model is pre-converted to the ONNX format in advance on the master node before distribution to computing nodes. This preliminary format conversion eliminates the need for time-consuming format transformation during container image distribution, accelerating the deployment process.
2Ease of operation
If multiple Internet-of-Things devices download container images from a centralized repository, then image distribution can be managed centrally, but network bandwidth resources are competed for resulting in network congestion and delayed container starting
Solution Approach 1:
The patent segments the centralized image distribution system into multiple distributed computing nodes that can independently serve IoT devices. Each computing node can download and cache images locally, reducing the bandwidth burden on the centralized repository and enabling parallel distribution to multiple devices simultaneously.
Solution Approach 2:
The patent introduces computing nodes as intermediary layers between the centralized repository and IoT devices. These intermediaries cache and distribute images locally, reducing direct traffic to the centralized repository and improving overall distribution efficiency through local content delivery.
Data Source
AI summary
The present disclosure discloses an Internet-of-Things-oriented machine learning container image download system and a method. The Internet-of-Things-oriented machine learning container image download system includes a master node and a plurality of computing nodes; the master node is configured to store and convert a machine learning model, and build a machine learning container image from the format-converted machine learning model; and issue an image download instruction to each of the computing nodes after image information of the machine learning container image is completely built; and each of the computing nodes is configured to receive the image download instruction, download the machine learning container image, and start a machine learning container; and receive data collected by Internet-of-Things devices, and return a data processing result to the Internet-of-Things devices. According to the present disclosure, the image distribution efficiency of the machine learning container in an Internet-of-Things-oriented scenario can be increased.