Hi Ugur,
first off, I have to admit that I did not (yet) make any such performance/scalability tests. That being said, the overall intent of LWM2M is to
be very resource friendly in the first place so that constrained devices can be managed having limited memory, cpu, bandwidth resources. So on the server side, we are trying to keep as little state a possible/necessary to be able to manage the devices. Thus,
even if I do not have numbers verified by tests, I am pretty positive that you should be able to connect way more than 1000 devices to a single leshan instance. I would think that 100000 up to a million should be easily possible on an instance. One of the
important factors influencing the number of the managed devices and performance should be the number of LWM2M Observations you create and maintain since their state is kept in memory. It seems that this is your intended use case because a device can only “upload”
data to the LWM2M server by means of a Notification sent in reply to an Observation created from the server side.
Maybe Julien can provide some information as well based on his experience of using leshan at Sierra Wireless?
In any case, if we are not able to manage several hundreds of thousands of devices per node (instance) we are doing something wrong and need to
address this in our future development.
Regards,
Kai