Documentation on the different values of graph.dataaccess?

[TippyToes]

Hello!
I’m reading through the docs on the github and I’ve realized that there isn’t any documentation on the different values of graph.dataaccess and that they mean. Why is RAM_STORE the best, and what are the other values and their usecases?

I perused the codebase and also found these threads on the matter:

Running Multiple Graphhopper servers specifically for different regions

So due to the size of files of larger regions, would it make any sense performance wise to run multiple instances of Graphhopper each with a different region’s .pbf file? I’m not too familiar with the devop side of things but would it even be possible to basically have someone access the API and request with a Lat,Lon and determine which region it’s in then use the specific instance of Graphhopper? It doesn’t sound that feasible but due to the amount of memory required to run the whole planet I…

Understanding store, RAM vs MMAP

Hello, I am trying understand how the stores are meant to be used. I have generated a graph (car - shortest) on the planet osm file. The graph was generated and contracted in memory. It’s then flushed on the disk (RAM_STORE). The ghraph’s files were then moved in another server that provides route calculation (using graphhoper-0.5) over JMS. So the graph can be read only but route calculation must be thread safe. If I load the whole graph in RAM everithing works like a charm. Problem comes…

But I’d like a definitive or concusive answer. Happy to add this to the docs if I’m answered.

[easbar]

There is some documentation here: graphhopper/core/src/main/java/com/graphhopper/storage/DAType.java at 116586191332ef94e4aceee5b93115af140bc096 · graphhopper/graphhopper · GitHub

The difference between RAM and MMAP is basically that the former uses memory from the Java heap while the latter uses memory outside of the heap. As long as there is sufficient memory the performance difference will be rather small (RAM usually being a bit faster), but if the memory is limited the JVM will crash if you use RAM while using MMAP the OS might start swapping data between the disk and memory making the process much slower but keeping it alive. Whether or not this is acceptable depends on what you are doing, but typically it is useful for data that is read rather seldomly.

[karussell]

There is also this blog post that describes this a bit.