Capacity Question

I have a need to place 7069 Nodes in one server. I only want to ping these nodes every 5 min, with no SNMP. Has anyone had this many nodes on one server or close to? I would welcome any official Solarwinds capacity recommendations.

Find more posts tagged with

Accepted answers

All comments

FormerMember

What else are you collecting on that poller? If it's just 7000 nodes, and you're just polling for status, that may work but then the question is how many interfaces are you polling, how often are you collecting statistics? It's all the same poller, so it all adds to the same pool. It sounds like you'd be pushing a single server.

aliendan

We would only be doing ICMP. No Stats or Interfaces. Is there a way other than entering the nodes as ICMP only to turn off all SNMP?

jtimes

7069 Nodes in one server. I only want to ping these nodes every 5 min

Never tried that...But my highest total elements on a single polling engine was around 5,900... and it was too much; I wanted to dig into the issue further, but it was losing/dropping to much SNMP data. My theory was a combination of demand on the SQL server (having to write so much data at once), disk queue lengths and/or also buffers on the polling engine's NICs getting flooded... but I never got that deep into it ...

Could I ask why you would want to just ping so many nodes? Are they already in your Orion as SNMP nodes as well?

aliendan

Yes, these are currently all the nodes in our production environment. They are spread across 4 pollers. Our production Data Base is on the company SQL servers and SVC. Both of which go down from time to time for maintenance and issues. The goal is to have a solution that would continue to monitor be it only up / down while the production system is down for maintenance or any unforeseen situation. The poller would also be located away from the main site.

FormerMember

Drop all interfaces. Then set all stats gathering to 600 minutes. You might get it all on one poller.

jtimes

Man that sounds like justification enough for you to get your own dedicated Orion SQL server for your current production Orion implementation. Burning a polling engine in your current Orion won't solve this; you are looking at deploying a seperate Orion SLX with it's own DB (kind of costly...)

I could be "offbase" but; here are some alternative idea's:

1. Get two (or more) copies of the Standard Edition Toolkit, put them on seperate PC's in your NOC and load half the nodes per pc in the NPM, run as needed, and only keep the data for 24 hours in the pc's local SQLExpress db...

2. I don't like it, but... use a free utily like Zabix or netmon, both of which are free, well sort of they both have a limitation on number of nodes..

Network_Guru

ICMP polling only at 300 second intervals should be OK for 7000 nodes (as a back-up solution).
I was running over 6000 elements with 99% polled with SNMP on a single server install.
It was slow but manageable (used 3 disk Raid 0 for DB).