I recently discussed how hyperconverged solutions are all about simplifying data centre management. Thinking about the difficulty of adding simple unified management to existing products illuminated the Nutanix Acropolis move. To get a really hyperconverged solution, you need to remove all of the legacy components and have a total platform that is designed for simplified management.
There is no doubt that data centre virtualization was an enabler for hyperconverged infrastructure (HCI). The pain of data centre virtualization is also the reason for the creation of HCI products. So it is natural that the first big HCI products use the same hypervisor that brought virtualization to a vast number of data centres. It is a big enough transition for some customers to give up their old storage arrays and LUNs when they put in HCI. Asking them to swap out the hypervisor at the same time might be a step (or three) too far. So the path to market uses the incumbent hypervisor. Once you are established, there might be an option to replace the hypervisor too, but not at first.
HCI product release dates are measured against VMware product release dates. How many weeks after the release of vSphere 6 was it before Nutanix released its support? The lag is inevitable to allow testing of the final software release and develop the non-intrusive upgrade processes of which Nutanix is justly proud. But this lag is a source of displeasure for customers. Early adopters want to be on the latest versions of everything right away. Often new releases fix bugs that impact production VMs. Any delay between hypervisor release and support is a bad thing for an HCI. Reducing or eliminating release lag is highly desirable for HCI vendors and customers.
One of the common feedback items for Nutanix is that there is no integration between the VMware management tools and the Nutanix Prism console. I cannot deploy or change the hardware configuration of a vSphere VM from Prism. Nor can I see or manage the Nutanix backup and replication of a VM from vCenter. It is a fail on the objective of simplifying data centre management when I must use two consoles for managing one VM. A vSphere client plugin would have been a great solution, and it is the solution that SimpliVity chose. The challenge comes when VMware decides to make changes, again. A recent VMware change was to deprecate the vSphere client and prefer the web client. Now SimpliVity needs to develop a new integration plugin. This is a fairly normal thing for VMware to do, to change the integration points for third-party products. It results in more effort for its partners, and I imagine some frustration too. I am still surprised that Nutanix does not offer an integrated console to manage both its HCI and vSphere parts. Customers are left with two consoles to manage their VMs.
To cure both the lag and the dual console problem, Nutanix chose to offer an option of its own hypervisor. Having its own hypervisor means that it controls the release cycle. As a result, there is no lag between its hypervisor release and HCI support. The hypervisor update doesn’t get released until it is supported by the HCI. This also means that it owns the integration APIs and the management tool. One management tool can manage both the hypervisor and the VMs. This is the simplification that HCI should offer to customers.
It makes a lot of sense for Nutanix to keep simplifying data centre management by having its own hypervisor. That doesn’t mean that customers are all going to choose to use the Acropolis hypervisor. Ultimately, simplified infrastructure management is about not having to spend so much money and effort on infrastructure. The money and effort can be spent elsewhere to improve the business. But a risk has a cost. Introducing a new hypervisor and management tool is a risk. If the cost of the risk is higher than the savings or return, then you don’t take the risk. The Nutanix Acropolis hypervisor will be attractive to customers who feel the risk is worth the savings.
One type of customer for Acropolis will be existing service providers. Often they will already be using the KVM hypervisor that Acropolis is based on, so the risk is lower. These same customers will be hurting from the difficulty of managing KVM. They will appreciate the value that Nutanix’s management brings. These customers also do not like to use a lot of manual processes and consoles. They develop their own web portals to allow their customers to self-service. The service providers care about a rich API that they can call from their own portals. Nutanix’s approach of having an API for everything first, then a GUI and a command line, appeals to service providers. This is also a piece of the data centre market that Nutanix would not get into with a VMware or Microsoft-based offering.