Why does it take so long to load stops using Network Analyst?

Question

I'm using the network analyst tool in arcpy to take breadcrumb trails and approximate routes using a network dataset. I'm wondering why the process of loading the stops into Network Analyst takes so long. The input data can include up to 10,000 points, which are only loading at about 15-20 stops/second. Every other step of the process (pre-processing, building network dataset, solving for routes) is relatively fast. Does anyone have any idea how I could improve performance?

MikeBrown5 · Answer

I don't understand why, but when I add file geodatabase containing the network dataset instead of an enterprise geodatabase cut my processing time from 45 minutes down to 5 minutes. Not my ideal solution as this adds a lot of extra steps and lines of code, but it will work for now. I would still like to find a solution that allows me to skip this step with similar results.

JaySandhu · Answer

Most likely the spatial index on the streets feature class  that makes up the network dataset (or the points you are loading) is missing or stale. See below for information on how to update it:

Rebuild a spatial index—ArcGIS Help | ArcGIS Desktop

Melinda Morang · Answer

Loading stops is doing more than just copying records into the Network Analysis layer.  It's actually "locating" those points on the network, meaning that it has to determine which network edge or junction is the closest one and record that location.  This process can take some time, particularly with large numbers of input points.

If you have a large number of inputs and are going to use them many times in different analyses, the efficient way to handle the slowness is to first use the Calculate Locations tool to pre-calculate these network locations in advance.  It saves the network location fields to the points layer.  Then, when you load the points into any Network Analysis layer, you can choose to use the existing network location fields.  In that case, it will load much faster because it doesn't need to perform the calculation again.  It really is just copying the data and nothing else.  However, if you make edits to your network dataset or decide to use different network restrictions, you will need to re-calculate your network location fields as they should be considered "stale" in that case.

Another, less common, reason why loading locations may be slow is if your network dataset or  your input points are missing a spatial index.