I have been using pgRouting for some accessibility analysis to various facilities on the network and experimenting with different ways of making the process faster.
My initial network had 28000 edges and to solve a catchment area problem for one location on the network to all other nodes on the network was taking 40 minutes on a 2.93GHz quad core processor with 4GB RAM (Windows 7 PostgreSQL 9.2 PostGIS 2.0.3 and pgRouting 1.0.7). I put the query into a looping function that processed the facilities in order but any more than 4 and the machine would run out of memory as the complete solution is stored in RAM until the loop finishes.
First step, reduce the number of edges in the network to 23000 and number of nodes to 17000 by removing pedestrian walkways, alleys, private and restricted roads. Now the query is solved in about 12-14 minutes using about 200MB RAM per facility. Continue reading Speeding up pgRouting