You can probably ignore the servers. Calculate how many ports are on the switches, and you'll perhaps get a rough idea of the requirements.

As you can now fit 24 cores in a single system relatively inexpensively, we can scale pretty far on a single system  as our poller runs multiple parallel processes. :)

Splitting MySQL to a separate server with fewer, faster cores helps. Storing RRDs on SSD (or ramdisk if you can afford the RAM) is very important too.

adam.

On 2014-07-11 01:37, Daniel Josue Rodriguez Paiz wrote:

Thanks for your reply, the equipments are switches and servers, is for
the monitoring of a ISP so is a large network.

Thank you very much for your reply.

2014-07-10 18:03 GMT-06:00 Adam Armstrong <adama@memetic.org>:

On 2014-07-11 00:52, Daniel Josue Rodriguez Paiz wrote:

Hello,

I'm interested on using Observium Pro to monitoring a large network of
15,000 devices more or less.

Does anyone know if observium works fine in a large network?

Thanks in advance for your replies.

Hi,

Observium scales primarily on I/O throughput for writing RRDs, which scale mostly by port numbers on network devices. (servers scale more by device number).

15,000 is a very large number, are they switches, cpe, servers?

I don't think we have any installations scaling that high, I think the largest I know of is ~3000 devices, mostly switches, so a very large number of ports.

adam.
_______________________________________________
observium mailing list
observium@observium.org

http://postman.memetic.org/cgi-bin/mailman/listinfo/observium [1]



Links:
------
[1] http://postman.memetic.org/cgi-bin/mailman/listinfo/observium

_______________________________________________
observium mailing list
observium@observium.org
http://postman.memetic.org/cgi-bin/mailman/listinfo/observium

_______________________________________________
observium mailing list
observium@observium.org
http://postman.memetic.org/cgi-bin/mailman/listinfo/observium