Hi,
does anyone know how many messages one gateway can handle at the same time? I mean even if it’s a private network it should still proceed the messages coming in from not assigned devices in order to find out if its devices from the private network. I didn’t found anything in the docs. Does anyone know something about it?
Regards
Maybe this document can give you an answer:
or “Known and Unknown Facts of LoRa: Experiences from a Large Scale Measurement Study” by JANSEN C. LIANDO, AMALINDA GAMAGE, AGUSTINUS W. TENGOURTIUS, and MO
LI, School of Computer Science and Engineering, Nanyang Technological University, Singapore
Unfortunately, I do not get any specific information from these documents. The results of the simulation do not reflect how many messages a gateway could theoretically be limited to at the same time. Of course, other factors play a role, such as the computing power of the gateways, etc. But I am also specifically concerned with the Kerlink gateways, which I may not have mentioned before.
Max ‘at the same time’ concurrently will be limited to the number of channels that a gateway supports, which is usually 8. Packets will not all arrive at the same time however.
It will be more useful to specify your use case and ask questions related to that.
OK, so I’ll put it back in to the Kerlink category - any more information, as @bluejedi suggests, you want to share so we don’t all spin our wheels guessing.
Fundamentally most gateways are the same as they are all based on one of the Semtech chipsets and the gateway isn’t doing anything particularly tricky - that’s all in the Semtech chips. All it has to do is take the uplinks it is given and send it via some transport mechanism. So not really just Kerlink.
You could use a spreadsheet to model an environment but you are more likely to lose uplinks due to overlapping transmissions than have the gateway overtaxed. So concentrate on ensuring the devices have some level of randomness about it’s precise transmit time, keep the power down and the DR low to keep transmissions short both in time & range and deploy more gateways just far enough apart.
If a spreadsheet model is too much, use the back on an envelope:
1,000 devices, four times an hour, 42 bytes at DR4 is 200ms, so in a fifteen minute period you have 200 seconds of air time over the 900 seconds over the 15 * 60 * 8 = 7200 channel/seconds available. Assume 50% inaccuracy and you still have 3600 channel/seconds available - so we are using 25% of the capacity, leaving plenty of room for other devices to send uplinks that aren’t on TTN. This is pure Bistromathics, but gives you some idea of what is possible.
Eight isn’t a bad number for modeling, or realistically lower given that randomness will never lead to an ideal.
However, in actuality the orthogonality of spreading factors permits two messages on the same frequency at different spreading factors to be received at the same time.
A traditional (sx1301/sx1308) gateway has in addition to 8 variable-SF sub-IF frequencyes, 8 decoders; these are not hardwired to the channel frequencies, but rather dynamically assigned to whatever looks interesting, which at any given time might include multiple packet possibilities on one channel, and none on most of the others.
An sx1302 gateway has 16 decoders; again, dynamically assigned. So in theory it could better leverage the possibility of having two packets at different spreading factors on the same frequency at the same time, by being less likely in a busier situation to run out of decoder instances to process them.
(The sx1302 is a smart buy anyway, as being made in a more modern silicon process even with greater number of decoders it consumes much less power in total leading to a reduction in the notorious tendency of the earlier versions to run hot and being more suitable for barely-possible alternate-power installations)