Address Locator taking forever to complete in ArcGIS Pro/ArcMap

[DavidBondy1]

I'm trying to build a address locator in ArcGIS Pro 2.3.3 and I've been experiencing issues with the length of time it takes to complete the address locator. I've built one using the same source data and it took about 20 minutes to complete. When I've run it with the adjusted data, it still hasn't completed after 21 hours.

 I needed to make adjustments to the dataset to standardize some of the values between the municipalities so I built a FME model to handle the data transformations. The source data was the same for both which is open data collected from 21 different municipalities, totaling ~2,800,000 records for the original and new, standardized datasets. The locator syle is 'US Address - Single House Subaddress' and has 1 primary table, 1 alias table, and 1 city alias table.

When I run the Create Address Locator tool against the standardized data, it continues increasing the number of cached records without moving past that stage. When i  stopped it after 21 hours, it was at ~13,000,000,000 records cached.

Does anyone have any insight into what may be contributing to this?

EDIT: I ended up finding the source of the problem. My Alternate City Name table had a bunch of duplicates in it which I'm assuming was causing the address locator tool to generate A LOT of additional relationships that weren't necessary, I refined the tables based on unique values and the address locator is now being created is a reasonable time frame. 

Thanks to those of you who provided feedback.

[JoeBorgione]

At the risk of sounding like a Parrot, can you update to 2.5 and also try the new style Create Locators?  Also, where do the source data layers reside?  Often times working across a network can torpedo geo-processing especially with that many features.

[DavidBondy1]

Hey Joe, thanks for the reply. 

Unfortunately I can't update my version of pro due to limitations with the license server. I did run the data against the Gazeteer - Single Field style, which completed successfully. So i'm thinking now it may be an issue with the number of populated cells which is taking longer for the pro to build and index. during the transformation every field for street type and direction was populated by splitting apart street name.

The source data is on my local drive. So I should be getting the best performance I can.

[Robert_LeClair]

David - kind of thinking out loud (in my home office), but are you overwriting the existing *.loc or deleting the old one and creating a new one?  How large is your GP history?  You may want to delete that.  It's probably not related but one never knows - clear out your AGP Display cache - mine was 8.5 GB!  Also are you creating the *.loc from *.gdb feature classes or SDE feature classes?

[DavidBondy1]

Hey Robert,

I'm not overwriting an existing .loc file. I'm creating a new locator in a new folder each time. I'll definitely try clearing my cache, good idea.

[ShanaBritt]

I would agree with Joe in suggesting to use the Create Locator tool in ArcGIS Pro 2.3.3 to create a locator based on the Point Address role to build your locator with the 2.8M features. These locators are supported on ArcGIS Server 10.6.1 or later. There are a couple of geocoding patches for ArcGIS Server (Enterprise) 10.6.1 and 10.7.

Regardless of the tool used to create the locator, when creating a locator with reference data that contains millions of features, it is necessary to have at least 3 to 4 times the size of the data in free disk space on the drive containing your temp directory because files used to build the locator are written to this location before the locator is copied to the output location. If you do not have enough disk space, the tool will fail at some point during execution when it runs out of space. Also, keep in mind that when creating very large locators, you should have a machine with enough RAM to handle large memory-intensive processes.

When you created the locator with the Create Address Locator tool using the US Address - Single House Subaddress locator style did you have suggestions enabled? If it was enabled and you disable it, does the time it takes to build the locator improve?