Hello Community!
hope you're well.
I'm evaluating new network monitorign solution in my company, and one of candidates for this is obviously Zabbix.
we have done some very initial checks of functionalities vs our requirements, so we would like to check what would be hardware requirements for our deployment.
our setup is like this:
- 2 DCs
- 4500 network devices monitored (total).
- current tool polls 110K metrics per minute (SNMP)
- 2K of total device number are cisco switches/routers which have extra data polling via Cisco DNA API
+
as addon (we do not at the moment) we would like to directly monitor 5k AP with Ping + few SNMP metrics (we do monitor via WLCs but it has some drawbacks)
we have data pollers/collectors configured with 2 clusters (1 per DC).
if single DC VMs are shut down, then other is to handle all the monitoring load.
in regards of DB - according to company requirements we have to use company provided postgres, so no need to handle this load..
Could you please advise how many core per DC and many RAM should i expect to handle this kind of load?
Is it better to have 1 zabbix server per dc + few minor proxies, or maybe one uber server?
I have no need to distribute proxies accross multiple network zones.
Thank you in advance!
hope you're well.
I'm evaluating new network monitorign solution in my company, and one of candidates for this is obviously Zabbix.
we have done some very initial checks of functionalities vs our requirements, so we would like to check what would be hardware requirements for our deployment.
our setup is like this:
- 2 DCs
- 4500 network devices monitored (total).
- current tool polls 110K metrics per minute (SNMP)
- 2K of total device number are cisco switches/routers which have extra data polling via Cisco DNA API
+
as addon (we do not at the moment) we would like to directly monitor 5k AP with Ping + few SNMP metrics (we do monitor via WLCs but it has some drawbacks)
we have data pollers/collectors configured with 2 clusters (1 per DC).
if single DC VMs are shut down, then other is to handle all the monitoring load.
in regards of DB - according to company requirements we have to use company provided postgres, so no need to handle this load..
Could you please advise how many core per DC and many RAM should i expect to handle this kind of load?
Is it better to have 1 zabbix server per dc + few minor proxies, or maybe one uber server?
I have no need to distribute proxies accross multiple network zones.
Thank you in advance!
A proxy is good for 6-700 nvps.. sometimes more.. depending, how you build it, what DB you use etc.. But considering the amount of devices... Start with ... maybe 10-12 of them (5-6 proxies per DC)... In our case, 4cpu, 12G memory. It can run 2 proxies at the time, if half of cluster is dead...sqlite3 db on /dev/shm.. You may not need to run 2 per host, if you have some backup hosts in cluster.. Might get away with smaller hosts then...
Comment