Address locator takes 20x longer to rebuild in Pro 3.3 than 3.0... why?

[AllenDailey1]

I've discovered that, since upgrading to Pro 3.3.2 from 3.0.3, rebuilding address locators takes about 20 minutes, whereas before, it took about 1 minute. 

This is the case both with a Python script and when doing it manually.  It even happens when rebuilding an address locator I created from scratch in Pro 3.3.2 (as opposed to the existing locators I had made with Pro 3.0.3).

This worries me because I'm wondering if it means something is wrong...?

Does anyone know why this is happening?  Has anyone else experienced this behavior?

Thank you!

Allen

[MichaelVolz]

Since Pro 3.4 has now been released and available for a few weeks, have you tested whether it takes 20 minutes there as well?  I would suggest on a test machine.

[AllenDailey1]

Thanks for your reply.  Well, since we have to stay on Server 11.3 for the foreseeable future, we don't want to use Pro 3.4 yet, since 3.3. is the one that's officially compatible with Server 11.3.  But I can keep that idea in mind in case we want to at least test it out, like you said, just in case it provides any interesting information.

[Robert_LeClair]

@AllenDailey1 - do you have any schema changes prior to rebuilding the address locator?  From the internals:  use the Rebuild Address Locator (Geocoding) tool for updating the locator with the current reference data and use the Create Locator (Geocoding) tool to re-build the locator when a schema change (e.g., adding a field) is made to the reference data feature classes.

[AllenDailey1]

Thank you.  No, there haven't been any schema changes.  
I did use those tools to create and rebuild my locator.

[JoeHeggenstaller]

We recently upgraded our Pro environment to 3.4 and discovered the same issue. Address Point Locator rebuilds used to take 1-2 minutes. Now it's taking 23 minutes. No schema changes to reference dataset. Created a brand new 3.4 locator for testing and still extremely slow to rebuild. Would appreciate someone from the geocoding team to look into this issue. Thank you.

[AllenDailey1]

Thanks for sharing that this happened to you too!  Interesting.

Here's how I solved the problem.  I checked the specs on the server where our scripts run, and it had way below the recommended RAM for ArcGIS Pro.  We upgraded the RAM, as well as the processor, and rebuilding address locators is back to taking only a short time.  It's still a little longer than with Pro 3.0, but maybe like twice as long rather than 20 times as long.  My guess is that rebuilding address locators consumes way more resources with the newer version of Pro; I don't know why.

However, a caveat is that before discovering this hardware issue, we had tested rebuilding locators in Pro (rather than a script) on other computers, all of which had okay hardware for Pro.  But the rebuild still took a long time... so it's still kind of mysterious.  I don't know if it needs a very particular processor or what, or if it's faster via script than via Pro interface.

[ShanaBritt]

@JoeHeggenstaller  Can you provide more details about your data and environment?

1. How large is the data used to build the locator?

2. Where is the locator being saved to? Local or network location?

     a. If local, is it to an SSD drive?

     b. If local, how much free disk space is on the machine you are building the locator on?

     c. Where is the data stored used to build the locator?

3. What was the field mapping that is used to build the locator?

[JoeHeggenstaller]

Hi Shana,

I'll provide the information you requested, but the issue has resolved itself with an upgrade to ArcGIS Pro 3.4.2. The locators now take less than 30 seconds to rebuild, which is comparable to what it used to be, prior to the 3.4 upgrade.

The point file dataset has more than 110K records and is stored in a SQL EGDB. The locator is stored on a network share of a virtual server.

I won't bore you with the field mappings, since there really isn't an issue with the data, but appears that it was an issue with the version of the software.

Thank you for trying to help!!

[ChristopherGraff]

@JoeHeggenstaller Any hints at what the problem was? We're at 3.3 and can't upgrade yet. We're setup very similar to you and see the same issue. We have half the records you do and it takes 30+ minutes now.