Hi,

Well with that huge number of nodes I would say 5 minutes (you didn’t say how much more than 5mins however) is already really good, generally I would expect a lot longer..

Assuming we are talking about NodeSet2 format XMLs and normal loading to the server, this might not be the best design in general, as that would load them into memory and if you have that much nodes then usually at some point running out of memory starts to be problem. A database might be better in this case.

You might need to build your server the MyBigNodeManager way (see the SampleConsoleServer example), then the data would not be as UaNodes and is basically up to you how you will load or store it. Still if the input to your application is the NodeSet2 file with a million nodes even just processing that might take a while..

Note, if that was done with 3.x of the SDK, you could try with 4.x, it might be slightly faster (only a bit and even then depending on the number of tags in the model), but with that number of nodes it might do some difference.

If you can, this could help (if you are not going the MyBigNodeManager way):
Check that you have defined each Reference only once in the model file. SDK automatically creates the reference to both directions even with one definition. If it is defined in both directions in the model, then SDK needs to do an extra step per reference when loading. This step is directly related to the number of references the node already has.