Monitor app performance using Performance Profile

AppSheet maintains a performance profile for every app. The performance profile is similar to the Audit History log, recording each sync, add, update, or delete recorded by the AppSheet backend. Along with each entry, there is detailed performance information captured during the execution of the operation. 

To monitor app performance:

1. Open your app in the editor.
2. Go to the Manage > Monitor tab and expand Performance Profile.
3. Click Launch performance analyzer.
   
   The Performance Analyzer shows:
   • A list of recommendations for improving performance, if applicable
   • Summary graphs showing:
     • Average duration (seconds) per operation type
     • Average virtual column computation time (seconds) by app version
     For example:
     
     
   • Audit History log results based on filters selected (listed below)

4. Filter the log by:

   • Syncs, Adds, Updates, Deletes, API calls, or Documents

   • Start or end date

   • Failures only

   • Table name, user ID, or rule name

5. Within a record, click the Performance icon to view performance details.

6. Optionally, click Download Search results to download a copy of the search results as a JSON payload.

By default, the Performance Profile is shown for a limited recent period of time. Richer filtering and analysis capabilities are available as part of the Enterprise Plus pricing plans.

When analyzing the performance data, look for:

• Tables that take a long time to read. Perhaps they have too much data?
• Virtual column expressions that take a long time to read. Perhaps they can be written more optimally?

Most often, a long sync time is due to the number and size of tables. Consider the following:

• If you 10 tables in your app, AppSheet's server has to read them all from the cloud backend. AppSheet will only read a few of them at a time in parallel. For example, accounts with self-service plans will only have 2 or 3 threads loading tables in parallel. Business subscriptions will have 4 or 5 threads loading tables in parallel.
• If a table is very large or takes a long time to load, then it slows down the overall sync time.
• Often there are references between tables and, therefore, automatically added REF_ROWS() virtual columns that compute the relationships between the tables. It may appear that one of the virtual column computations is taking a long time to complete. However, this is rarely the case. The time required by this computation is typically used to fetch the table itself before computing the relationship. In other words, fetching tables is almost always the source of delay.
  
  Note: If you write your own virtual column expressions with inefficient SELECT() expressions, this may cause delays in sync time.