question about snmp check performance

[mconigliaro]

it looks like im going to have to do some bulk-loading of snmp mib values in order to monitor individual ports on a switch. since its quite possible that i'll have to monitor switches with 100 or more ports, i wonder if i should import directly into my default Host.SNMP template or if i should make a seperate template specifically for switches.

the reason i ask is because im somewhat new to snmp, and i dont know how efficient (or innefficient) it would be to individually check 100 snmp values on every host that uses the default Host.SNMP template. Any thoughts?

[mconigliaro]

anyone have any thoughts on this?

[Alexei]

How many hosts you've got? What is expected total number of SNMP checks?

[mconigliaro]

well, let me explain what im trying to do...

i currently have several host-level templates (eg: host.windows, host.linux, host.snmp) and several application-level templates (eg: app.msexchange, app.backupexec, etc.). my goal is to be as generic as possible in each of the templates, so that zabbix does as few unnecessary checks as possible. this also makes it easy for the people here to add a new device to zabbix.

my host.snmp template currently contains icmpping and sysUptime, since those are pretty much guaranteed to exist on any snmp-enabled device. i also want to track the ifDescr[*], ifInOctets[*], ifOutOctets[*], and ifOperStatus[*] of each interface on the device. the only problem is that if we're monitoring a switch, we may have to monitor 100 interfaces. if i add 100 interfaces to the generic snmp device, then ill be doing ~400 unnecessary checks on things like routers that only have 2 or 3 interfaces. im not sure if thats a big deal in the snmp world, but i get the impression that it will eventually become a big deal if i add enough devices to zabbix.

the more i think about it, the more i think i should just create a seperate 'switch' device to handle devices with large numbers of interfaces.