Audit Logging CPU Impact

This topic is only applicable to on-premises installations.

Each user transaction generates multiple log items. The Audit Logging feature enables you to perform deep analysis and monitoring of activity within Archer. This feature affects the environment sizing due to increased traffic over a TCP/IP or UDP port. The final destination for this traffic is a monitoring tool.

The following table provides an estimate for the number of items logged based on typical user activity, based on a single user.
Type of Transaction Approximate Number of Monitoring Tool Entries

Data Feed / Data Import

4 times the number of items in source file

Logging into default workplace

240

Switching workplace

150

Display All

35

Advanced Search

40

View or Edit a Record

50

Lookup and Add a Cross Reference

55

Apply/Save

95

Delete a Record

60

Execute a Search or Report

25

Modify Report

10

Save Report

55

Export Report

40

Print Record

15

Access Quick Links

35

Administrative Functions

50-60*

*Average. Creating a simple questionnaire has a log entry rate of about 500, but most administrative functions are much smaller. Since few users typically perform administrative functions, their impact is limited.

Audit Logging can potentially generate a large amount of network traffic between the servers and the monitoring tool. Depending on retention policies, additional disk space for storing log data can be required. Consult the sizing documentation for the chosen monitoring tool to predict this storage requirement based on an analysis of expected activities.

Audit Logging has no impact on the Database Server, but increases CPU utilization on the Web and Services Servers. In a large environment with hundreds of concurrent users, CPU utilization could double. If Audit Logging is required for a deployment, factor this into the hardware design and acquire additional cores to ensure performance is acceptable. If the increased CPU usage does not cause contention (for example, when adequate cores are available), end-user page load times are not impacted.