just uptime and overall bandwidth consumption

[velma41abbott]

Hello,

We are working on a project to provide internet access via WiFi hotspots in refugee camps, using consumer routers and access points, all running OpenWrt firmware. Currently, we have around 5000 units deployed and need an open-source solution to monitor their uptime and WAN bandwidth usage.

The metrics we need to track are minimal—just uptime and overall bandwidth consumption—so we are considering using Zabbix for monitoring. Would a single server with 16GB of RAM and an i5 or i7 CPU be sufficient to handle the load of monitoring 5000 devices?

Additionally, what would be the most efficient method for monitoring these access points? Should we use SNMP, Zabbix Agent Passive, or Zabbix Agent Active for gathering the required data? We’re looking for the most cost-effective and scalable solution to handle this large deployment.

Any recommendations or advice regarding best practices, performance considerations, or alternative tools would be greatly appreciated. Thanks in advance! este enlace

[SitecKris]

To compare, I had 2800 PLC's on an oilrig in a stand-alone Zabbix, running on an Intel NUC with I3 quadcore, 12GB ram and 500GB disk.
I just needed metrics of throughput and TCP errors, so I just polled them using SNMP once every 120 minutes. Never had a problem.

I later Implemented Proxies to do the polling (500 devices each) using RaspberryPi v4. Just so I could scale faster and dont be to reliant on one server.

The data wasn't important, so I dident care about HA/Redundancy.

Sometimes you dont get the money and time to do the perfect solution, and you gotta go with what works and what you can maintain.

[markfree]

I'm always amazed at how reliable Raspberry Pis have been. I myself have a Pi 4 running 24/7 with a small Samsung flash without a single hardware issue for over 4 years.
So I would consider some Pis for your setup, especially the newer Pi 5.

Considering the number of hosts, a distributed setup seems more appropriate.
Fortunately, Zabbix is very flexible and can handle different monitoring methods. It depends on the monitored device.

Thinking about scalability and low cost, maybe some Pis as proxies at the deployment site and a pair of remote Servers should be cost-effective enough.