Performance and Concurrency

Some of the enrichment steps that this service runs are CPU and memory intensive, as you would expect for advanced machine learning steps.

Therefore, you will need to configure the concurrency on the REST enrichment step within your pipeline so that you do not send too many concurrent requests. If you do not do this, then you will get timeout errors and the documents will be marked as ‘error’ status.

This configuration is covered in the ‘Communicating with the Python Rest Service’ section. A short cycle of empirical testing will establish the optimum value for this.