Slow geocoding with GDB based StreetMap files

Question

ESRI used to provide the StreetMap Premium files in two versions (SDC and GBD files).  I avoided the using the GDB based files because geocoding is slower than with the SDC version.  Now that I am forced to use the GDB version I find that geocoding is painfully slow.  On my system, geocoding 13 million addresses with the SDC based files took about 5 to six hours.  The time is double that when using the GDB based files.  Is there some kind of fix for this or settings that can be changed to optimize the geocoding speed?

Jim_Gay · Answer

Michael,

I do not have a solution for you but I do have a question.  I need to setup an automated process to geocode about 2.5 million addresses for an initial run, after that the runs will be much smaller but I need this to be an automatic process.

I've opened a discussion around this topic:

https://community.esri.com/message/459609?sr=search&searchId=8b1ea69d-1e76-4630-a59b-773dcf2a25df&searchIndex=3#459609

If you could share some information on how you are accomplishing your geocoding that would be great.

Thanks!

MichaelFitzwater · Answer

The address records and locators are both on the same drive.  The locators, standard StreetMap premium locators, are free standing and not in a Geofile database.  Also, RAM and CPU have never been a problem.  It's not often that I need to geocode that many records at one time, once or twice a year.  However, the slowdown applies to any size file.  The SDC of the StreetMap files could geocode up to two million records per hour.  The GDB version of the StreetMap files file runs at between 500,000 and one million records per hour.  It appears to me that the Geofile database is, for whatever reason, a much less efficient database than the SDC database.

JoeBorgione · Answer

Only for California...  8th largest economy on the planet: might as well be nationwide...

Seems like there have been other posts over the years with respect to large (13 million is large) geocoding. One suggestion I've seen is to have have the records and the matching data on the same drive, instead of across a network.  If you have your locator as a database object, I suggest deleting it and rebuilding it stand alone in the same drive as your records and data.

There must be millions of road segments if not tens of millions, right?  That's a whole lot of work to do.  Have you watched what your computer is doing?  Is the cpu and or ram maxed out?