Limits on “Unlimited” modules

robertcbrowning

We’re running a primary polling engine, four auxiliary polling engines and a dedicated web server to a separate physical SQL2005 server. All the polling engines are running on VMware platforms and are performing well. We will be deploying additional servers soon, but it takes time to get approval!  We currently track about 20000 NPM elements and 2000 APM components. We’re running NPM 9.5 SP4 with APM modules loaded across all of the engines. Additionally NCM, NTA & VoIP are loaded on the primary polling engine and the web server. All product is SLX/ALX/DLX

We are looking at continuing the expansion and need to better understand the “unlimited” limits so that I can pre-arrange budget for when it is needed. My management do not like shocks!

 

NPM

I realise that physical polling engine servers top out around 8000 elements while VMs should be limited to 7000 and that adding more elements can be achieved by installing another auxiliary polling engine.

 

APM

I understand that the recommended count is 2000 per ALX instance, but what is an instance? I have APM polling engine modules on every one of my NPM polling engines and they are set as Poller Bound, but does this count as one instance or multiple instances. The Admin > APM Settings > Licence Summary only shows the total count and the fact that it is “unlimited”. Is it 2000 per auxiliary polling engine or 2000 per primary NPM/APM. If this counts as only one instance, how do I expand the APM component polling?

 

NTA

I understand that the recommendation for NetFlow is a max of around 400 source interfaces per NetFlow instance. But I only see my primary polling engine listed as a NetFlow Collector. How can I distribute collectors across my auxiliary polling engines to expand this limit?

 

Web

As more users access Orion from more locations, it could be useful to deploy a second additional web server. Is this possible? 

 

I realise that I could deploy multiple primary engines and then consolidate with the EOC, but that seems an expensive and over–complex solution.  Being able to expand by just adding another engine has to be the simplest and most scalable option.

 

I’d appreciate any guidance on scalability of these modules.

Many thanks.

Robert

chris.lapoint

Hi Robert, thanks for the post and very well thought out questions.  Please see below for responses to each of your questions.

NPM I realise that physical polling engine servers top out around 8000 elements while VMs should be limited to 7000 and that adding more elements can be achieved by installing another auxiliary polling engine.

That's correct.  We have NPM customers with well over 50,000 elements based on the use of additional pollers.

APM I understand that the recommended count is 2000 per ALX instance, but what is an instance? I have APM polling engine modules on every one of my NPM polling engines and they are set as Poller Bound, but does this count as one instance or multiple instances. The Admin > APM Settings > Licence Summary only shows the total count and the fact that it is “unlimited”. Is it 2000 per auxiliary polling engine or 2000 per primary NPM/APM. If this counts as only one instance, how do I expand the APM component polling?

I think there might be some misunderstanding here.   A single APM 3.1 server can handle up to about 7000-8000 component monitors, assuming 10-15 monitors per node.   You can get around 20,000 component monitors (max for single APM instance) by adding additional pollers.

NTA I understand that the recommendation for NetFlow is a max of around 400 source interfaces per NetFlow instance. But I only see my primary polling engine listed as a NetFlow Collector. How can I distribute collectors across my auxiliary polling engines to expand this limit?

Collectors won’t do much to expand the limit because, due the voluminous nature of NetFlow data, the bottleneck is the database.   To boost performance and throughput, you’ll need to look at Top Talker Optimization and other tuning options – see here:  http://support.solarwinds.com/kbase/ProblemDetail.cfm?ID=1296.  For example, with Top Talker Optimization enabled, you can achieve 60,000 flows per second in peak with a single NTA instance.    Beyond this, you can deploy another NPM/NTA instance and use EOC for rollup.

Web As more users access Orion from more locations, it could be useful to deploy a second additional web server. Is this possible?

You can actually add as many additional web servers as you want.  This may help some, but the most significant gains in web responsiveness can be achieved by ensuring the database is optimized for maximum performance.  For example, here are some recommendations for physical SQL server:

• Dual Quad Core Processor 3 GHz or better
• 32-64GB RAM (64bit you can increase this. More memory the better)
• Hardware RAID Controller with a battery backed-up write back cache.
• Disk subsystem with as many spindles as budget allows. Configured in a RAID 1+0 array to allow for maximum write performance.
• Minimum of 10 disks.
• 15k RPM Drives (higher speed for optimal performance read/writes)
• Log vs. data – separate them onto separate controllers if possible – if not, separate channels at least
• Connectivity – speed is key

HTH,

-Orion Product Team