Introduction
This function is only applied to the internal communication of AI Inference Server, without any external connector communication.
This function allows you to fine-tune whether data loss is allowed/disallowed and to configure the size of the data point queue.
The tuning is currently performed using the Industrial Edge Management System and "AppConfig".
The configuration file can update either during installation of the AI Inference Server or with the Industrial Edge Management
configuration function. (Update configuration applies only to pipelines that are not being executed. Values of the executed pipeline cannot be changed).
Configuration format
AI Inference Server can operate without using App Config.
The internal format of the configuration is expected in JSON and accepts the following keys:
stopPipelineAtDataLoss: boolean(default: false)
Configures whether the currently executed pipeline is to be stopped immediately after the first deleted data point.
This applies both to the orchestratorInternalQueueSize and the runtimeInternalQueueSize
orchestratorInternalQueueSize: integer(defaults: with TimeShifting: 1000 With DirectConnection:10)
If the pipeline configuration includes the activation of DirectConnection, the value of orchestratorInternalQueueSize is ignored.
Configures the internal queue size of the internal component called the Orchestrator, which passes data points to the Python Runtime. The default value depends on whether or not TimeShifting has been activated. If TimeShifting is activated, DirectConnection is deactivated.
runtimeInternalQueueSize: integer(defaults: with TimeShifting: 1000 With DirectConnection: 10)
If the pipeline configuration contains DirectConnection activated and does not have multiple steps, this value is ignored.
Configures the internal queue size of the internal component called Python Runtime, which processes the data points and passes predictions to various connectors.
Configuration example
{
"stopPipelineAtDataLoss": true,
"orchestratorInternalQueueSize": 100,
"runtimeInternalQueueSize": 10
}