|
those 10,000 objects in the database are most likely just 'stubs' that get cleaned out when you do the ovtopofix -a. They do not impact netmon since they are not polled. They are created by NetView when he reads arp tables but when NetView checks out the device he sees through your seed file that he should not discover them, but he does not remove the stub. That is normal behavior. It does not impact your polling. It only adds a little to the memory requirements of ovwdb but not to the performance.
100 ms and 180 ms is a pretty long time....and that is just latency. Get an overloaded switch/router and your snmp requests take even longer. Are you getting any 'false downs' followed by 'interface up' traps right afterwards? If so, then change your timeout to a little longer.
I would have some pointers for tuning if you were running on Linux or AIX, but I don't have experience running large networks from Windows platform. I can just offer some general knowledge. I would recommend using ICMP status instead of SNMP.....unless you have to use SNMP for a type of node. Especially if you are managing a global network since ICMP is lighter. I would turn off the ping spray if you haven't already.
Hopefully someone else running the size of environment can give you some additional pointers.
As someone else said...yes, all your devices should be in DNS. Especially the interface that you are sending SNMP requests too.
Stephen Hochstetler shochste@us.ibm.com
International Technical Support Organization at IBM
Office - 512-838-6198 (t/l 678) FAX - 512-838-6931
http://www.redbooks.ibm.com
|