Enable Feedback Chaining

You must enable feedback chaining by making some configuration changes on your downstream Media Server.

In the [Chaining] section of the configuration file, set the parameter RemoteAnalysisConfigurations. Specify a comma-separated list of configurations that can be used to perform remote analysis for other Media Servers, and the maximum number of processing sessions to allow concurrently for each configuration. For example:


This example permits Media Server to use the configuration files RemoteFaceRecognition.cfg and RemoteObjectDetection.cfg to perform analysis for other Media Servers.

An upstream Media Server could run two process actions that include remote face recognition, and three that include remote object recognition, and the downstream Media Server could handle requests from all of these actions simultaneously.

An upstream Media Server could run more than two process actions that include remote face recognition and more than three that include remote object recognition. In this case the requests for remote analysis might have to be queued on the downstream Media Server until previous requests have completed. However, the remote analysis tasks on the upstream server are unlikely to send every video frame downstream. Remote analysis requests that originated from one upstream process action can be processed while there are no requests being received from another. Each remote analysis request contains only a single record, so even if requests do need to be queued, the requests do not wait in the queue for very long.

The limit for the total number of connections to the downstream server is specified by the parameter MaxRemoteAnalysisConnections. The value that you choose for this parameter is the maximum number of process actions, that include remote analysis, that you can run concurrently on the upstream server(s).

You should also ensure that the downstream Media Server requests sufficient visual channels from your License Server. Media Server requires a visual channel for each remote analysis request that is processed concurrently. For the previous example, you would need to set VisualChannels to a value no less than 5. The visual channels are required regardless of whether Media Server receives any requests.

The configuration file for the downstream Media Server might therefore contain the following:


[Chaining] RemoteAnalysisConfigurations=RemoteFaceRecognition:2,RemoteObjectDetection:3 MaxRemoteAnalysisConnections=10

The chaining configuration parameters MaxProcessingSessions and QueueTimeout do not apply to feedback chaining.