Hello, we are facing a project for deploying a nation wide 915-923 MHz LoraWAN network with around 5000 nodes connected to it.
Has anyone used Lora Server Project for deploying large networks like that? What’s your experience? What you guys think could be the biggest challenge(s)?
And more important… is the Lora Server Project the right choice?
Thanks to anyone willing to help me figure this out.
Cheers
I currently have a network running 52,000 active end-devices. A total of 17 different app works in 5 different organizations.
Loraserver is adequate and stable. but don’t using mosquitto. having serious ram problems. hangs continuously. we decided to use emqx after doing various tests. an awesome mqtt broker. cluster support is very nice and practical.
You’re welcome, mqtt broker is independent of lora-server. lora-server cannot see which broker is using it.
The official docker container https://hub.docker.com/r/emqx/emqx
Operating systems Ubuntu
-lora-server on dedicated server (not docker)
-lora-app-server on docker on another server
-emqx 3 different locations on a dedicated servers (not a docker) (as a cluster)
-We don’t record data, that’s the truth. customer wants to register itself. Since they are mostly state institutions, they are recorded on mssql.
I don’t remember much detail. We investigated it in 2017.
Mosquitto was starting to cause trouble every day. very serious zombies were forming stacks (in ram) and becoming unstable. there was no internal software watchdog. couldn’t lift yourself off. Cron and I restarted daily. A few weeks later the final number of devices reached 15,000, and communication became very intense. more trouble started several times a day.
I’ve created a team, explored all the alternatives that are open source and support clustering.
all options were stress tested. emqx was very successful. when you start using it you already love it, it is user-friendly and very stable software.
if you have any other applications to recommend, please let me know. success is to stay up to date. I would like to examine immediately.
As a point of comparision, we’re running all LoRa Server components, along with Redis, in Kubernetes. PostgreSQL is in RDS. We are running VerneMQ in clustered mode, also in Kubernetes.
I can’t speak highly enough of VerneMQ, it is a great project, both for horizontal scalability and the latency requirements that LoRa conversations need compared to your average bulk MQTT traffic.
Unfortunately, there is no correct answer to this question.
The missing packet is a little. There are different reasons for this. For example, a water meter brand in Turkey, sending messages uncorfirmed. And unfortunately this constant, can not be changed. When the ADR is at SF7-8, some messages are lost due to changing physical conditions.
The number of devices per gateway is very variable.