HCI market consolidation with the fall of HyperFlex
In this blog
If you haven't already heard, Cisco announced an official partnership with Nutanix, and less than a week later, announced that HyperFlex will no longer be a product. We don't yet have all the details on how this impacts existing customers and options for a path forward, but we do know a lot about how we got to here, and that's important in itself.
Below is a quick history of HyperFlex, with some expert perspective on why this ultimately came to be and where this leaves the HCI market.
When Cisco is brought up in a conversation, most people think of "networking things" or "compute" in the form of Unified Computing System (UCS). Over the years, Cisco has attempted to grab storage market share either through the acquisition of Whiptail (which we do not speak of) and more recently, in 2017, the acquisition of SpringPath. These software acquisitions were then integrated into the Cisco UCS ecosystem in an attempt to give that ever-elusive, single-pane-of-glass management for storage and compute resources.
SpringPath, the underlying storage management layer of HyperFlex, was founded in 2012 by former VMware executives and later sold to Cisco for $320 million. The timing of the acquisition was no coincidence. A few months earlier, HPE had announced the acquisition of SimpliVity for $650 million, which at the time left Cisco on the outside of the HCI market looking in with no answer.
VMware's vSAN and Nutanix were leading the pack in the HCI space. Dell's majority ownership of VMware (before selling it off) gave them the inside track to tightly integrate with vSAN in the form of VxRail, and now HPE had an offering with SimpliVity. The acquisition of SpringPath gave Cisco a path forward, but left them playing catch-up.
After acquiring SpringPath, Cisco began integrating its software with UCS, but faced another challenge. A year before the acquisition, hybrid nodes were the preferred option, offering SSD caching for fast writes without the high cost of an all-flash solution. However, as is often the case with technology, all-flash storage became increasingly affordable, and the benefits of lower latency and more performance outweighed the small cost increase over a hybrid solution.
Being privy to Cisco's roadmap while providing prioritization feedback to the product management team, we strongly urged Cisco to re-prioritize feature development and pull support for all-flash configurations forward. The majority of HyperFlex nodes that were sold quickly became all-flash configurations, but the time to bring the solution to market further pushed Cisco behind the curve when it came to software features.
Without getting into too many of the details, HyperFlex had some shortcomings that others had already figured out. Some of these are listed below:
- No replication.
- No native or external integration for disaster recovery automation.
- No erasure coding.
- No file services.
- No single-file-restore capability.
- Required fabric interconnects, increasing costs.
Over time, Cisco addressed some of the above deficiencies. In an attempt to differentiate itself, Cisco also began developing a specialized ASIC for accelerated storage efficiency, which had a promising future if the roadmap was executed successfully.
Cisco's focus on Intersight and HyperFlex management shifted resources away from developing or augmenting new features for HyperFlex, which lagged the technology further behind and made it difficult to compete with other HCI solutions. The time spent on these efforts allowed the competition to pull further ahead, addressing customer needs and market trends.
There are other factors that could be listed, but there are really four major events that led to the "final nails" for HyperFlex:
- Gartner's definition of HCI changed to require multi-hardware support. This left Cisco off the official HCI quadrant. A knee-jerk reaction to integrate with other hardware platforms pushed feature development further back.
- A not-so-known-to-many-effort within Cisco to offer HyperFlex as a backup appliance/target (against recommendations).
- The decision to stop developing Intersight Workload Engine (IWE) – a KVM-based hypervisor. The initial use case was to be used for Kubernetes with the end goal of also being a virtual machine hypervisor.
- Broadcom's acquisition of VMware (still pending at the time of publication).
We could dive into each one of these but the biggest event that led to the demise of HyperFlex, in our opinion, is Broadcom. We mentioned that HPE had acquired SimpliVity a few months before Cisco's acquisition of SpringPath. A few years later, HPE also announced a partnership with Nutanix. Not only was the partnership an OEM offering, but also a GreenLake solution running on Nutanix AHV. As luck would have it, that partnership would prove to be potentially vital in a post-VMware acquisition world. More on this below.
No one could have foreseen the acquisition of VMware but the timing of the new partnership between Cisco and Nutanix, on the surface, appears very reactionary to potential future market demands. So why was this a problem for Cisco after the announcement was made?
HyperFlex supports vSphere and Hyper-V, but most deployments are done with vSphere as the hypervisor. Microsoft is positioning Hyper-V improvements through Azure Stack HCI/Hub, which could pose a problem for Cisco with HyperFlex and Hyper-V support.
Cisco's relationship with VMware has had its ups and downs, and Broadcom being more-or-less a direct competitor of Cisco, the acquisition introduced complexities. With Broadcom owning VMware, vSAN (VMware's SDS offering) is not a place Cisco wants to invest time and money for integration.
Dell has PowerFlex, which they can push if the VMware relationship breaks down and starts affecting VxRail. PowerFlex only addresses the storage angle and not the hypervisor, but it does offer the flexibility of also being a pure software-defined storage technology that can be exported to just about anything.
Lenovo and HPE both have strong offerings with Nutanix and HPE also offers GreenLake, all of which support Nutanix's hypervisor AHV. This addresses both the SDS layer and gives customers an option to evaluate VMware alternatives, running on their respective hardware.
Given the above, Cisco is once again on the outside looking in, trying to address industry concerns surrounding the acquisition of VMware, which HyperFlex could not do – questionable relationship with Broadcom/VMware, no pure SDS play to capture storage and no answer for an alternate hypervisor.
All of this said, HyperFlex filled a need for customers and had its place in the market. The execution and priorities were possibly misaligned and a few things outside of Cisco's control forced their hand.
Oh, HyperFlex…you will be missed (probably).
So where does this leave the HCI industry and who is left that are perceived as viable options for large enterprises?
- VMware vSAN Ready Nodes and VxRail
- For customers who are not concerned about the post-acquisition VMware landscape, vSAN or VxRail are great technologies.
- Nutanix (running on multiple hardware platforms).
- Nutanix's hardware flexibility allows customers to have flexibility not only in hardware, but also hypervisors as they arguably offer the closest thing to a vSphere equivalent.
- Dell's PowerFlex
- PowerFlex has an HCI offering that runs on vSphere, but can also be utilized to present storage to external (potentially non-VMware) hypervisors if required.
- Azure Stack HCI
- Multi-vendor hardware integration that helps customers with an Azure public cloud strategy bring cloud-like features on-premises.
There are other technologies that either are HCI or consider themselves to be that are not listed above. The keyword from the list is based on what we feel fits the needs of large enterprises. This is of course a debatable topic.
As mentioned at the beginning, there are a lot of details still missing but we're expecting those to start flowing in the very near future (if not already by the time of publication). Look for an update from the WWT team with those details.
All in all, the reality is that HyperFlex will cease to exist within the next few years leaving customers with many questions. If you're one of those customers with HyperFlex in your data center, please do not hesitate to reach out to your account team and we can help you navigate the situation.