Configuration for driving distances performance

openrouteservice
1

Hello! We are trying to deploy OpenRouteService in Kubernetes for an OSM file of about 2GB.

We want to use ORS to get ONLY driving-car distances in miles between coordinates, using matrix API. We do not need any other info, so weare in the process of evaluating/minimizing the configuration needed for query performance and build time.

For services we only use “matrix” and “routing”
For routing the only profile needed is “driving-car”.

From methods “ch”, “lm”, “core”, are all of them needed for our purpose? Or “ch” is just enough?

2

Hey,

yes, ch should be enough, as matrix currently doesn’t use neither lm nor core.
So indeed, that should be the minimal config needed :slight_smile:

Best regards

3

Thanks you @jschnell

One more question.
Do I need to define any ext_storage for my case?
Will it improve the performance if it is added?

1 Like
4

Hey,

no, they’re used for any extra information such as way surfaces and waytypes being returned with the route, should you need that.
In your case, that’s not necessary, since you’re aiming for a minimal config :slight_smile:

Best regards

5

Thanks again for the promt help.

We are using -Xms5g -Xmx10g for this 2GB OSM file. I suppose it is fine.
Is there anything else we need to consider for performance in terms of config.json values?

As we plan the OpenRouteService on our K8s cluster to be heavily used, we are thinking of having a volume with the built graphs accessed by different pods based on autoscaling policy (due to load), with BUILD_GRAPHS set as False for all (So the very first pod to build the graphs on empty volume).
Is it a valid approach for performance?