Maximum number of elements per poller.

videotech

Maximum number of elements per poller.

How exactly is this calculated?
I read that the maximum is 6-8K and is dependping on the hardware of your Poller.
Also I assume that these figures are based on the default polling interval, which is 5 mins?
So if your polling interval is 2 mins, how does this effect the total number of elements, and
by how many?

Also the hardware that comes into play include CPU, RAM, etc. It does not mention dual core,
quad core CPU. Is Orion multi-threaded.??

How exactly do you know when you have reached your limit on a Poller?

I think Solarwinds really needs to clarify this exactly, maybe with a formula or something?

What do you guys think.

Thanks

FormerMember

 A formula sounds like a good idea, but we've found the number of variables involved to be very high.  Worse yet, those variables interact with each, which would mean that the resultant formula would be a multi-variate equation, and we would have to stop development for a while just to try to measure all the variables and produce an accurate formula.  Hence, no formula. 

Since that was probably an unsatisfying answer, I'll try to get your specific questions answered:

 

read that the maximum is 6-8K and is dependping on the hardware of your Poller.

Not really.  That's a typical scenario, but we have customers polling twice that number with very beefy hardware and very long polling intervals.  But for the majority of users with good hardware and default polling intervals, that's typical.

 

Also I assume that these figures are based on the default polling interval, which is 5 mins?

 Default polling interval is 2 min for status and 10 minutes for statistics. 

 

So if your polling interval is 2 mins, how does this effect the total number of elements, and by how many?

Orion's poller is single-threaded, so it can collect X bits of information per minute.  If you're asking it to poll more often, then you'll get fewer elements.  The exact ratio is hard to say.  If you try to poll really often, then you can run into issues where not everything gets polled before the next poll begins and you'll see gaps in data.  Other factors come into play such as database I/O efficiency, network latency, etc.  As a consequence, it's hard to predict in advance for a particular network.

Also the hardware that comes into play include CPU, RAM, etc. It does not mention dual core, quad core CPU. Is Orion multi-threaded.??

Single-threaded, so extra CPUs won't help much.  But RAM on your Orion server and on your SQL server matter, especially the latter.  As I mentioned above, the database I/O efficiency matters.  If your database is configured to optimize input speed, you'll get more out of it than if it's optimized for redundancy because for very large Orion installs (e.g., multiple pollers), the amount of data being written to the database per unit of time can get pretty high.  Most customers who actually hit that probably consult a DBA for configuring the database optimally.

How exactly do you know when you have reached your limit on a Poller?

The telltale sign is gaps in your data.  You'll see graphs that have chunks missing.  That indicates that the poller was not able to finish all polling in one cycle before it had to begin the next cycle.

 

So while I agree that a formula would be great, the reality of how truly variable our customer base is and how many factors impact Orion performance have combined to make such a formula elusive.  Until then, we have our rules of thumb which have proven quite effective (if not perfect).

That's my perspective, but I'm obviously part of SolarWinds and the actual users might have a different opinion which you (and I, actually) would like to hear.