SolarWinds Lab Topic Suggestion - Explanation of Stacked Polling Engines

stevenwhunt​  I know there's a lot that goes into # of elements vs processor speed/load, among other things. I think it should be that simple, if you think about the concept.

The easiest to do this would be to multiply polling load (L)x number of pollers on an individual server (P). So if you have 1 poller your polling load shows 20%, but if you 2 your polling load shows 20% x 2 pollers (20L x 2P) = 40% and it states (40%/200%). If you have 3 stacked you see 60% (20L x 3P) and it states 60%/300%. ETC.

If computing were only that simple.

The *only* problem I have with stacked pollers is that the only true ways you see it are in both licensing and polling load. Polling load should really be a number measured like % processor time or % idle time, so that if you had 3 polling engines your polling load could theoretically go up to 300% - like how %time is really %time times # of CPU's. That would make a lot more sense than the 1-100% as it's defined today. Thoughts, stevenwhunt​ / cobrien ?

All depends on the specifications on the host, really. If you have plenty of compute spare, and more importantly enough bandwidth on your polling network to cope with the additional traffic from the single host, then you could stack them. That said, if you have the OS licensing available and the compute resources free, deploying it on a separate server would be better. Sure, it'll make maintenance more of a pain when installing SW updates, as you'll have to update the APE separately, but it's the way to go if you resource isn't an issue.

Yes. I'm working on deploying our first APE, but wondering if i should create a second VM for it or just stack it on the same VM that our Primary one is on