Hyper-converged infrastructure (HCI) pioneer and market leader Nutanix has added the ability to connect with external storage from Pure Storage from its HCI stack. The move is Nutanix’s second overture towards external storage – the first was to Dell PowerFlex last year – and marks a further rupture with a strict hyper-converged infrastructure model
The announcement – made at this week’s .Next event in Washington DC – helps Nutanix position itself to take advantage of customer need to de-risk exposure to VMware commitments after its purchase by Broadcom and the increased costs and lowered deployment choices that are reported to have resulted.
The move to add external Pure Storage FlashArray – its most performant array product line based around Connectivity – capacity to Nutanix HCI allows existing VMware customers to consider moving to a different hypervisor for virtualised workloads. Nutanix offers its own Acropolis hypervisor but can run any other that the customer wants to deploy. That comes because Nutanix’s operating system – AOS – allows it.
Nutanix was a pioneer of so-called hyper-converged infrastructure, which saw compute and storage bundled together in nodes that could connect in grid-like fashion to form clusters, often with server and storage components scalable independently. This was a particularly attractive proposition to customers that lacked deep skillsets as they were relatively easily deployable and scalable.
And so the idea that Nutanix should allow the use of external storage from third parties somewhat goes against the original principle it pioneered. It does, however, make a compelling proposition.
On the one hand, it allows customers that want to swerve away from VMware to do so relatively easily into a hypervisor-agnostic environment. It also provides scalability in terms of capacity and performance that Nutanix may have previously lacked in its Pure hyper-converged infrastructure architecture. That’s possibly made more urgent by the current upsurge in AI workloads, which can run to very large scale in terms of data volumes.
Here, Nutanix marketing vice-president Lee Caswell explained the difference, comparing cluster sizes possible with external storage – which run to the hundreds – with Nutanix as it has usually been deployed.
“HCI users scale their clusters at much smaller sizes,” he said. “They’re typically six to eight, eight to 12 or 16 nodes. And so you can see that we’re bringing performance and scale of a storage system into Nutanix usability, the Nutanix usability model.”
Customers will be able to manage Pure capacity through the Nutanix Prism interface, from where they can run replication, snapshots and DR capabilities.
Caswell was also keen to highlight the independence of storage in the Nutanix architecture compared with VMware.
“AOS was architected independently, modularly, independent of the hypervisor,” he said. “That’s in stark contrast to V San, for example, which is integrated in the kernel of vSphere. You can’t run vSAN without vSphere. It’s part of vSAN. In fact, when vSphere ships, the vSAN bits are in it. From the start, we’ve been able to offer a choice of hypervisor.”
Does this all mark a retreat from hyper-converged for Nutanix? Not at all, said Caswell: “It’s not a replacement for HCI. We’re not walking back from any of the HCI value proposition that we have, and there’s still lots of customers. HCI is only 20% penetrated in the market today. At the same time, we’re recognising that storage and HCI will coexist, and that Nutanix is going to provide a common way to go and manage across storage and HCI systems.”
But why now? Caswell was keen to highlight that the rise of IP storage and the relative decline of Fibre Channel has made it less of a technical challenge to integrate external storage arrays to the Nutanix world.
“Historically, most external storage has been Fibre Channel,” he said. “And the technical lift to go do hardware compatibility with Fibre Channel, the switches, and with Fibre Channel external arrays, has been a really hard problem, but now it’s a solved problem. What we’re focusing on is IP-based storage. That reduces a whole slew of technical integration risk.”
He also highlighted the push to change that the Broadcom takeover of VMware had provided: “Most storage systems attach to vSphere, and there was no compelling move in a very slow-to-move, conservative market, to move away from vSphere. Until Broadcom. The Broadcom changes are now pushing customers to de-risk future exposure they have to Broadcom business practices.”
