To say that when NetApp announced their HCI solution back in June created a bit of a kerfuffle would certainly be an understatement.  There has been a lot of back and forth on whether the Compute/Storage node system should be classified as an HCI solution or just a CI solution – and while I find the comments from El Reg interesting and sometimes very entertaining to read I personally feel that the argument is null – meaning who necessarily cares if it’s HCI or CI, or how it’s packaged, or what the many definitions of HCI are – to me, as a customer, in the end all I care about is whether the hardware/software combination can solve a problem I have, or deliver on a project I need to complete.    I got a close look at the solution when Gabriel Chapman from NetApp Solidfire presented at a recent Tech Field Day Extra event at VMworld – so while those “wordy” battles ensue, let’s us take a closer a look at the technology behind this…

Just a disclaimer of sorts – every article, comment, rant or mention of Tech Field Day Extra at VMworld 2017 that you find here has been completed on my own merits. My flights and accommodations to/at the event were kindly paid for by Gestalt IT – and I received a conference pass to VMworld through VMUG – however I’m not required or obliged to return the favor in any way. 


So what’s it look like?

NetApp HCI from a physical standpoint looks much like all other HCI solutions we are used to – a 2U chassis with the ability to have to 4 nodes in each chassis – pretty standard.  However it’s what’s inside those nodes which differentiates the NetApp HCI solution from others.  Instead of each node providing both compute and storage, NetApp splits their nodes into individual functions – meaning a node can contain just compute, a custom NetApp solution or a node can be a storage node, an all flash SolidFire solution.  The base package, or starter package if you will, does have a minimum footprint which is 2 chassis containing 6 nodes (4 storage nodes and 2 compute nodes), with two empty slots available for expansion.  Once you meet this minimum requirement customers and mix and match nodes as they please – want 10 storage nodes and 10 compute nodes, no problem.  How about 4 storage nodes and 7 compute nodes – again, no issues as the minimum requirements of 4 storage and 2 compute are met.

The nodes are essentially all we need as well.  Meaning we don’t have the traditional VM or controller workload that sits on top of each node to pool resources – all of the functionality is built into the nodes themselves, allowing the customer to utilize the resources provided for their workloads.


As you can see this isn’t the same design as the traditional HCI solutions we are used to – and this is what causes most of discussion around NetApp HCI.  As I mentioned earlier though in the end a solution really needs to meet a customers needs, and NetApp HCI does this by designing their solution around three main objectives, Guaranteed Performance, Flexibility & Scale, and Automated Infrastructure.

Guaranteed Performance

gp SolidFire itself provides some pretty nifty functionality when it comes to QoS – and it’s this technology that drives the guaranteed performance benefits of the NetApp HCI solution.    By providing the ability to define and enforce minimum, maximum and burst settings on a per VM level, NetApp is able to dynamically allocate and manage storage performance on our workloads.  For example, we can define a certain workload to always have a basic number of IOPs, allowing it to burst to a higher level for those one-off, month end type situations.  In addition to that we can set a maximum number of IOPs a neighboring workload can consume, ensuring that it does not in fact steal the performance needed by other workloads.  What this does is allows us to take many different types of workloads and run them within the same infrastructure, with the confidence in knowing that we can guarantee the performance needed for each and every workload we run.

Flexibility & Scale

fs Due to the way the NetApp HCI solution is designed it provides much greater flexibility when it comes time to scale your solution – meaning we can scale just our compute nodes if compute is an area where we are constrained.  Or, say we need just more storage, we can simply just add more storage nodes into the pool – or, both compute and storage can be added to existing or new chassis to meet the demands of our workloads.  NetApp certainly believes that one size doesn’t nessessarily always fit all – and allows for that with their architectural designs when it comes to scale.  Also, NetApp believes that just because we have this shiny new piece of kit sitting in our datacenters we aren’t simply going to toss all of our older stuff.  That’s why the NetApp HCI solutions isn’t limited to just the compute nodes that sit inside the chassis – but in fact, any external solution that needs storage could, in essence, pull from the HCI storage nodes – you have some investment in some SolidFIre storage there – might as well leverage it!

Automated Infrastructure

am Most modern solutions we deploy within our datacenters today involve some sort of published API consumption as well as rely heavily on automation – and NetApp HCI is no different.  The initial deployment of the solution is all hinged up on the execution of something called the NetApp Deployment Engine (NDE).  This essentially takes in some credentials and IP information from the user and automates the installation of the Element OS on our storage nodes, VMware ESXi on our compute nodes, as well as provisions vCenter complete with the HCI plugin and configures our networks, datastores, and vCenter inventory items accordingly – all within roughly 45 minutes from start time.  As far as the ability to build your own automation around the solution NetApp has a very comprehensive API that they offer around the solution.  Interesting fact – the first three iterations of the SolidFire release came without a GUI, and nothing but a bunch of API calls.  So  yeah, if you need to automate anything you should be able to find some sort of hook into the NetApp HCI solution to do so.

In the end it was great to hear from Gabe and to hear about how NetApp is entering the HCI market.  The solution is certainly unique, but is also very flexible when it comes to scalability – anytime a company provides a solutions which offers choice when it comes to scale it is a good move in my books.  The war on whether this truly is an HCI solution though, in my opinion is really null and void – Certainly the analysts may care, the people who feed whether or not a company or solution is included on a magic quadrant may care – but myself, I don’t!  What I care about is finding solutions to my organizations technical problems – and NetApp HCI could certainly fit in certain areas – be that said, traditional, first gen HCI as they say can also solve certain needs.  So, call it HCI or just call it CI, but in the end NetApp has a nifty solution that independently scales and solves certain customer needs – throw in some Solidfire QoS and automation sauce on top of that and that certainly gets a check in my books!    If you want to learn more about NetApps HCI solution head on over to the TFD event site and watch the videos!