Form Recognition sizing and performance

Products

• Data Loss Prevention Enforce
• Data Loss Prevention
• Data Loss Prevention Enterprise Suite
• Data Loss Prevention Discover Suite
• Data Loss Prevention Core Package
• Data Loss Prevention Plus Suite

Issue/Introduction

This article contains sizing and performance information for the Form Recognition feature.

Cause

You are using Form Recognition in DLP, and want to know more about modifications to some of the relavant settings, such as those found in the online help topic:

Advanced server settings for Form Recognition (broadcom.com)

Resolution

Sizing for Form Recognition

Symantec recommends that you dedicate a number of threads to the Form Recognition thread pool that is half the number of physical cores on your detection server. For example, if you have four physical cores on your detection server, dedicate two threads to the Form Recognition process.

You can set the number of threads dedicated to the Form Recognition detection process on the Server/Detector Detail - Advanced Settings page, using the FormRecognition.OPENCV_THREADPOOL_SIZE setting.

Improving Form Recognition performance

You can have a maximum of 500 forms across all Form Recognition profiles in any given DLP deployment. For best performance, Symantec recommends creating fewer profiles that each contain a greater number of forms, rather than having many profiles that each contain fewer forms. For example, you will achieve better performance with five indexes containing 100 forms each than you will with ten indexes containing 50 forms each.