Traccar like a GPS listener / 3000 units up
That's not a problem. Traccar can easily handle much more than 3000 devices.
Hello Anton;
Thanks for you answer, we made testing and Traccar works very well like a listener.
What about a linux requirements to handle 3000 unit?
Assuming that average reporting frequency is about once a minute, you probably need a server with 8GB RAM, ideally with SSD drive.
We are using Ubuntu 16.04.2 Server, 8GB RAM, SSD in raid 10 (Max Throughput 1200MB/s) and 8 Cores CPU. Our experience is go with SSD not with SAS. We did a trial with SAS drives and it's difference between day and night, especially for Database queries with indexing. We using traccar as receiver only to port it to our transaction server which handles the API to 3d parties and Web Interface. We have 2000+ Units, the CPU load is around 25-35% with all services running on the background. Our memory usage is around 1.5 GB operational usage.
hello guys;
The last month I was still running performance testings, and I am very happy with Traccar.
I decided to implement 2 Ubuntu VPS with SSD (thanks @webidus for your suggest), the distribution was a VPS to the Database and another server to the Traccar app), however, I am also concerned about how to get more stability.
¿What do you think abouth implement Load balancing?
Regards
Fabián
I would actually recommend to have Traccar and database on the same server. It eliminates connection issues and delays between server and database. Traccar itself doesn't put much load on the storage, so there is no impact on the database performance.
As for load balancing, Traccar doesn't support horizontal scaling at the moment. There are a few thread on forum about it, if you want to know more details.
I agree with Anton, install Traccar and the database on the same server.
If you want more secure / stability what I can advise is, use 1 Server as Receiver in this case Traccar and
make your own "look up" script which "grab" the RAW data from the Traccar database into another Server. Use this server as your transaction server.
Our business rule is "5000" Units on 1 traccar server. So, if you hit the ceiling you add a new "traccar server" into your domain.
Our philosophy is that you don't want disturb the "receiver" too much. It's pure for collecting RAW data and "inhouse" add-ons in our case such as
street names from GIS/Shape Files and ODO accumulated (auto re-parse for old data comes in) etc.
We do our 'reverse geocoding' inhouse because we don't want to be dependent on 3rd party.
For the transactions it's better to-do it on a separate server. Main reason for this is 'security' / 'scalability'.
You want "separate" modules in your business process. If 1 "module" falls down the other process still works.
Our current architecture is:
2x Traccar Servers
1x Transaction Server
Hello guys;
Thanks for your advices, I going to do the changes (DB and Traccar on the same server) to running performance testing.
I will try to do "grab" the RAW data from the Traccar database to insert on another Server and use this server as your transaction server.
Hola Fabian,
Can you help me on implementing Traccar like a GPS listener and disable the web platform ?
Thank you
Hello there;
I'm interested on implementing Traccar like a GPS listener and disable the web platform. However, I have the concern if Traccar could work properly listening events of more than 3000 units.
Regards