How many Network Insight collectors do you need?
I'm not sure if I should be proud of this or not, but I have quite literally deployed vRealize Network Insight for somewhere around 100 customers over the past couple of years. It's pretty easy to understand why - Network Insight is an awesome product that helps you plan micro-segmentation, understand application dependencies, and can even discover your applications using machine learning now.
The most common question I get, is "How many collectors should I deploy in Network Insight?"
And that's why we're here today! This will be a quick post, but I think it's an important question to answer.
vRNI Architecture Overview
The first thing we should discuss before specifying a number of collectors, is how the vRNI architecture works.
The first major component of vRNI, is the platform. The platform is an OVA deployed inside your vCenter environment that is responsible for the web front-end. It's also responsible for storing and processing flows, displaying them, and doing all of the cool analytics and machine learning stuff.
Important Note: While you can deploy a "cluster" of vRNI platform OVAs, it is not for high availability, but rather for scale. If you were to lose vRNI platform VM #1 - your entire service would be down, even though you have additional vRNI platform VMs deployed.
The next major component of vRNI is - you guessed it - the collector. The collector is deployed as a VM, and is responsible for collecting data from your data sources. In 90% of deployments, "data sources" means collecting data from your vCenter instance. But keep in mind - a data source in vRNI could be anything, such as:
Cisco Switches (including ACI!)
F5 Load Balancers
Native AWS and Azure cloud instances
Standard NetFlow/sFlow from physical devices
VMC on AWS
This is just a short list, but if you want a full list of supported data sources in vRNI, you can check out this link.
Number of Collectors
Now that we know the collectors report data into the platform, we just need to know how many collectors we should deploy. Here's my recommendation which has served the 100s of my customers well:
Deploy one collector per physical location, of form factor "medium" - with the caveats below:
If you have more than 4,000 VMs, go to a Large form factor
If you have an environment with tons of silos, and isolated environments - having multiple collectors within a location may make sense. That's OK, because they're super easy to deploy and you don't pay anything additional to deploy them.
If you plan on collecting NetFlow from physical devices, you'll actually need a dedicated collector. So if you have one physical location but want to collect NetFlow AND vCenter data - you'll have two collectors total.
That's all! I hope that was helpful! If you'd like to learn more about Network Insight, including how to do a full deployment of it and integrate with vCenter - you can check out my vRealize Network Insight Fundamentals course.