Do Hierarchies work across feature classes, or only within?

[ChrisHiggins]

I am looking to create a network out of two separate road files - one consists of highways and another of local and arterial roads.  I want the highways to be favored for the route analysis so that users traverse their local roads to reach the highway network, but was confused about whether hierarchies work across feature classes in a network, or if they only work within a single feature class.  The help files seem to suggest the latter.

[PatrickStevens]

If your network is not very large, for example, multiple states wide or full North America, the hierarchy will not make much of a difference.

The purpose of hierarchy is to make solves faster for long distance routing.  Hierarchy exists because there are far too many small streets in very large datasets and searching all of them to find the most perfect route can be slow.  So, the route solver will begin filtering out "lower level" roads as the search progresses. 

If you want your routes to favor highways, then give your highways faster speeds / quicker travel times.  That is the more accurate way to proceed.  The reason someone would take a freeway in real life is that travel on the freeway is quicker. 

In conclusion, you probably don't need to worry about creating a hierarchy on your network.  Focus on accurate travel times and your routes will be better. I hope this helped.

[ChrisHiggins]

Thanks Patrick.  The only reason I was considering a hierarchy rather than speed was that we have detailed traffic data for the highway and arterial links, but would turn to local roads to connect various points to the highway and arterial network.  Speeds in the traffic data have the potential of being well below the local road network (they would be static, based on 40km/h or something), thats why we wanted the local roads to just serve as a funnel to the traffic data links.

I think the way around it is going to be just combining them into a single feature then creating a hierarchy based on the respective networks - gets the analysis to the detailed traffic data links as soon as possible.  Unless anyone has any other ideas.

[PatrickStevens]

Hierarchy will work across multiple source feature classes.  What is called the "logical network" is built based on all of the sources.  It doesn't matter which line sources make up your edges, in the logical network it is all one big edge type.

But, once you have built your hierarchy, you might not see differences between exact and hierarchical routes.

Perhaps another solution could be to use historical traffic to scale down the speeds of your local roads during high traffic times of the day.  Or to add global turn delays for intersections at the local road (to slow down for stop signs or traffic lights).  Not perfect, but those options are closer to how road speeds actually behave.