A version of this column was originally published on Diginomica as EMC-VMware, Nutanix Battle For Enterprise Virtualization Platform Supremacy
Vendors and analysis regularly wax enthusiastic about the cloud, and for good reason when you consider the phenomenal growth rates at AWS and Microsoft Azure. Yet most enterprise workloads still run on internal systems. Indeed, spending on public cloud remains but a tiny fraction of a $3+ trillion IT market. While cloud infrastructure may represent the future of IT, with a mere $7 billion spent on “true” private cloud, i.e. internal systems with self-service provisioning of shared, metered compute, network, storage and application services, it seems obvious that when 77% of recent survey respondents claim they have an internal cloud, what they’re really talking about is a virtual server farm most likely running VMware ESXi, not anything that remotely resembles AWS. That’s why the latest product announcements from the EMC-VMware VCE consortium and pesky hyperconverged competitor Nutanix are so significant: they represent the battle for customers and mindshare as enterprises build the foundation for next-generation VM infrastructure.
Nutanix arguably invented a market and product category with its scale out, hyperconverged appliances combining compute, storage, hypervisor and virtualization management software in a convenient, integrated package. While Nutanix has plenty of smaller competitors, it eventually attracted the attention of enterprise virtualization’s big kahuna when VMware introduced the EVO:RAIL product line in 2014 and was later joined by Federation partner EMC with its VSPEX Blue products last winter. Both promised the same appliance-like simplicity, density and software integration as Nutanix with the advantage of using the same management stack and vendor support structure many large enterprises were already comfortable with. It is a classic contrast between disruptive innovator and established incumbent. Where Nutanix had the lead on features, flexibility and depth of product line, VMware/EMC had the benefits of early access to native VMware software enhancements, established business relationships and a large, global sales force.
Where both stumbled was pricing. With typical prices for a 4-node box well into six-figures, neither were particularly attractive to enterprises already loaded with racks of Dell or HP servers and both were off limits to smaller enterprises. Furthermore, as relatively new platforms, the evolution of performance and feature enhancements and scalability of hyperconverged products was unknown.
In separate, almost simultaneous, but almost certainly not coincidental announcements, VCE and Nutanix tackled each of these shortcomings. Nutanix led with a major software update that it claims “delivers up to 4x performance improvement for any workload with no additional hardware or software license” via 25 new software features. Although typical improvements are more like 2-3x, it’s still an impressive achievement particularly when you consider that a two-box, 8-node cluster of the all-flash NX-9000 product can now deliver up to a million IOPS, a figure once the domain of 7-figure all-flash disk arrays.
The more impressive half of the Nutanix announcement, particularly for enterprises using public clouds or multiple Nutanix sites for production workloads, is news that the Acropolis App Mobility Fabric can switch workloads from vSphere to the native Acropolis hypervisor “in minutes with minimal disruption and risk”, a feature that allows automatic failover and DR from a site running vSphere to another that doesn’t. It’s significant since not only does such hypervisor agnosticism threaten VMware’s hegemony by simplifying workload migration to Nutanix clusters from native vSphere, but it conceivably means the ability to migrate even complex VMware-based applications to AWS or Google Cloud using software from Nutanix partner Ravello Systems. Ravello software allows nesting one hypervisor, in this case, ESXi or the Nutanix Community Edition, inside another, for example Xen or KVM running on AWS or Google Cloud. However, by first migrating on-premise ESXi instances to the Nutanix hypervisor, organizations avoid the need to install VMware software or buy licenses before spinning up VMs on the public cloud, saving time and money. Ravello estimates it can provision workloads on AWS or Google using the Nutanix hypervisor “for less than $1 per hour.”
VCE VxRail Details
Where Nutanix addressed the price/performance shortcomings of hyperconverged systems by improving the denominator, EMC-VMware, aka VCE squarely target the numerator. In a briefing, VCE President and former head of EMC’s Systems Engineering Chad Sakac said the company has learned that the hyperconverged market is quite sensitive to entry-level pricing and that its existing systems (e.g. vBlocks, VxRACK, EVO:RAIL) were much too large, expensive and inflexible. In rebooting its hyperconverged line, Sakac said VCE focused on four parameters:
- Pricing: By starting small and scaling up, VxRail at $60K is less than half the price of entry EVO:RAIL products at $140-200K.
- Product: Optimized around flash storage (although available in both all-flash or hybrid configurations), VxRail adds storage features enterprises demand like deduplication, compression, erasure coding, remote replication and cloud object storage.
- Positioning: Forget the Swiss Army Knife approach and instead focus small hyperconverged systems on SMBs, departments within large enterprises and distributed, scale out workloads and not replacing large, integrated converged systems like vBlocks for traditional large enterprise workloads. As Sakac puts it, “Anyone who says that one converged or hyper-converged offer can cover every use case is (IMO) as high as a kite, or suffering from ‘single product delusion’.”
- Packaging: Appliance simplicity with a single sales and support channel for the entire hardware and software stack.
The result is VxRail, a hyperconverged appliance optimized for vSphere users comfortable using server-side storage (in the form of the just-refreshed VSAN 6.2) instead of a central SAN array. Using the same 4-node, 2U configuration popularized by Nutanix, VxRail packs up to 20 cores, 512GB RAM and 10TB of hybrid (flash cache)storage into each node with the ability to scale up to 16 boxes (64-nodes) per managed cluster.
Read between the lines of the VxRail PR blitz and it’s clear EMC and VMware have Nutanix, among others, in their sights. Indeed, the slide deck used to brief analysts on the product takes a shot at someone that sure resembles Nutanix in a chart showing the “most widely deployed hyper-converged solution in the market” showing VMware passing the “#2 HCI Vendor” last year and touting 200% YoY revenue growth in Q4’15.
Nutanix and VCE are chasing a multi-billion dollar opportunity. According to IDC, the market for hyperconverged systems, which it defines as products that “collapse core storage and compute functionality into a single, highly virtualized solution”, is growing at over 150% annually with a $1.1B run rate. Sakac says that VCE exited 2015 with a $3+B run rate, although the bulk of its business is still in converged (integrated) infrastructure using external storage arrays. Nutanix booked $241M last fiscal year, running at about $350M annually as of it’s October 2015 quarter, meaning it had about 25% of the hyperconverged market using IDC’s totals. The remaining 75% plus triple-digit growth leaves plenty for VCE, Nutanix and others to fight over.
The significance of the competition over hyperconverged is less about the hardware sales and more about the platform. VCE wants to serve VMware customers and keep them bound to its virtualization-cum-cloud stack. While Nutanix addresses those same VMware users, it is less concerned about keeping them locked into a specific cloud stack as long as they continue using Nutanix platform: whether the workload runs in vSphere/vCloud, Xen, KVM (OpenStack), or potentially even Azure Stack is less material. It will be interesting to watch how customers react and whether choose to optimize hardware around and existing and familiar software platform or prefer hardware flexible enough to host their cloud platform of choice.