How much we can scale Solarwinds NPM?

Orion can scale to this size and beyond without issue. There are many customers doing this today.

Do we have any case study? showing how many devices were being monitored, what kind of architecture and how many pollers were used? Any application performance issues observed.

Case studies can be found at the link below.

Resource Center | SolarWinds

One follow up question - Handling such big environments require multiple pollers. Is there a limit on additional pollers that we can add to single instance by Orion?

Can we have 10-15 additional pollers added to single instance?

Maximum of 100k elements per primary SolarWinds NPM server (i.e. 1 NPM server + 9 APEs).

https://support.solarwinds.com/Success_Center/Network_Performance_Monitor_(NPM)/Maximum_capacity_of_Solarwinds_infrastructure

You can spread the load across as many as 20 additional polling engines if so desired.

Thank you for this informatio, this helps a lot.

so, max of 20 additional polli engine with top limit of 100k elements one 1 instance of Solarwin - is this correct?

what does element count includes - Is SAM license count part of element count? I am assuming interface, volumes and nodes makes elemen count that we see on polling engine page.

I think it would be better to engage SolarWinds sales and speak directly with a Sales Engineer who will be able to walk you through Orion deployment architectures and discuss scalability limits. I will state that SAM components do not count against element scale. You could for instance, monitor 100,000 elements and 150,000 SAM components with a single Orion instance.

You should talk with SW SE's about NAM.  I'm planning on going that way, since it's cheaper than the alternative.  I'll end up with another poller (I already have four), and they'll throw in UDT, VNQM, HA, and IPAM (I already have NPM, NCM, NTA, and NTM).

My support costs actually will decrease $1,000.00 per year once I have NAM.  And it'll scale to 25 pollers (I think), spread across multiple diverse geographical regions, and they'll all report to one instance of NPM.

The trick is to remember that NAM supports up to 100,000 elements, and a single element is anything you want to monitor or manage.  I want to manage 50 Cisco 4510's, each with 384 copper ports.  Every port is an element, so that just ate up 19,200 elements right there.  And each chassis has additional elements like SVI's and port-channels/channel groups, Loopback interfaces, managed VLAN's, unmanaged VLAN's, CPU, memory, etc.

Worse, Cisco started adding MIB's and OID's for "controlled" and "uncontrolled" virtual interface information for every single physical port on a 4510.  If I wanted to monitor all of those, in additional the physical interfaces, I'd be at just about 60,000 elements for 50 chasses.  Thank goodness I'm not interested in monitoring those!

So I currently have about 800 Cisco switches, routers, firewalls, and UCS chasses.  I'm at the 12,000 limit of elements on two pollers in my city, and I've had to start monitoring local elements with other pollers I have in other states.  It works fine, but it does increase the "apparent" recorded latency, and it sends more control traffic across the WAN.  I'd rather have a third poller in my city, just to keep the latency stats accurate and the WAN utilization down.

If you end up with more than 100,000 elements, you just buy another instance of NPM and the right amount of pollers for the elements, and then add in EOC so you can see all of your data in one NPM-aggregated-view. 

See this link and Patrick's video about NAM:

Network Automation Tools | SolarWinds

Maximum 12k element on each polling engines but recommend 10k elements on each polling engine