Handling memory allocation with pgrouting in centrality analysis

[ricardo_rs]

Hi @cvvergara and @rnanclares,

Thank you so much for your quick response!

@cvvergara

In pgRouting in develop branch, There is the new function pgr_betweennessCentrality. Its the only centrality measurement that we have. Don’t know if its the one you are looking for.

I can try the pgr_betweennessCentrality to compare the results. We are using the edge betwenness centrality, but instead of considering all possible pairs of nodes in the network, we created a set of origin and destination. In any case, the new function seems totally relevant for the analysis! thank you very much. I will see how I can use it (I still have a lot to learn)

@cvvergara

A thing to mention is the way pgRouting 3.6.2 (3.6.3 soon to be released) and under works:
data on the database

Then, It sounds to me that it will be an important new feature. That’s great, looking forward for the 3.6.3 release.

@cvvergara

To be able to test your work.

Can you give the information in the following form:

• A database dump of the tables involved
• A query script that execute the different sizes on that database
  (aka for me is just a command to get the database working and ready to test)

Sharing the data using a database dump sounds now to me as the best option. I was not aware. Please, find here the link with the data. There are two files:

• A database dump of the tables involved: osgeo_pgrouting_dump.sql
• A query script that execute the different sizes on that database performance_quries.sql

@rnanclares

Using the bbox method is not a sound strategy, because in some cases the best route might fall outside of your bbox.

Very well observed. It is indeed a limitation. Probably there is a way to add some distance to the boundingbox (line 252) depending on the scale. For example + 0.01º .

@rnanclares

You forgot to mention how much RAM you’re using. I’ve previously worked with Brazil states and I didn’t have memory allocation issues.

I tried two options:

• Option A: Local machine with a total memory of 15G
• Option B: Server machine with a total memory of 47G

Again, thank you very much for your prompt and kind responses. Please, if there is a problem with any query or the database dump, let me know.