Is Nutanix a VMware copycat?

“When Nutanix has gone down stack to try to build their own hypervisor, across stack to build their own networking, and now going upstack to compete against Horizon—we view it as sort of a copycat, wannabe VMware move.”

–  VMware COO, Sanjay Poonen
CRN: Sanjay Poonen on ‘Copycat’ Nutanix and VMware’s
Independent Status with Dell EMC
.  09/19/2018.

Fifteen years ago, VMware introduced vCenter and vMotion to the world. If you weren’t working in IT in those days, it’s hard to grasp the impact of this now standard virtualization feature. It was like magic. Most (seasoned) IT folks today can still remember the circumstances of what they were doing when they first saw a vMotion demonstration. In my case, I decided to launch a channel partner business focused on VMware virtualization.

VMware did more than change the career trajectories of millions of channel partner employees and customer administrators. It also escalated the storage industry as organizations across the globe scurried to buy SANs. The driver? Page 37 of the original VMware VirtualCenter 1.0 User Manual’s VMotion requirements started off with the bullet point: All hosts must share a storage area network.

 

VMware continues to fire on all cylinders. Its revenues for 2018 were just under $8B, up 33% from the year before. Its market cap is now $64B (EMC purchased VMware in 2003 for $635M), and VMware is considered the crown jewel in the Dell portfolio. Fortune
just named VMware CEO, Pat Gelsinger, as the #2 Businessperson of the Year.

I continued evangelizing VMware virtualization even after selling my partner business to Presidio in 2008. I wrote many blog posts about the advantages of VMware, and the company posted a video of me on its web site talking about why vSphere had a much lower TCO than Hyper-V. I am honored to be one of a very small number of virtualization advocates to have been awarded the title of vExpert for all 10 years of the program.

I’m not sure I would have joined Nutanix in early 2013 if I had known it would eventually come out with its own virtualization platform. But despite my years of loyalty and admiration for VMware, time marches on. It is now appropriate to reflect on the day and night difference between the virtualization architectures of VMware and Nutanix.

With VMware, virtualization always has been, and continues to be, its focus. It continues to sell its proprietary hypervisor as both a stand-alone and bundled product, and ties its other solutions back to ESXi. Its VDI and Virtual SAN products, for example, require the VMware hypervisor.

Nutanix, on the other hand, follows the public cloud model of virtualization. The virtualization layer becomes invisible and exists as simply a feature of a software-defined infrastructure stack. Once in a while, a TCO customer will ask me if they get a discount if they don’t use Nutanix AHV. I chuckle because no one gets a discount for not using a feature.

In evaluating the virtualization approaches of both organizations, it is important to remember that Nutanix came to market 11 years after VMware. The times were very different both technically and economically. The contrast in virtualization architectures stems from each company’s heritage.

VMWARE – BORN IN THE WINDOWS ERA

Two years before the introduction of vCenter and vMotion, VMware launched its virtualization hypervisor, ESX. In those days, network connectivity was 100Mb, Intel processors only had one core and ran at 1.2Ghz. Memory was sold in 128MB DIMMs running at 133 MHz for $50 each. Flash was still a niche product.

VMware created the ESX architecture based on the resource and economic environment at the time. Going to market shortly after the dot com bubble burst, VMware’s emphasis, understandably, was on saving customers money. It enabled large resource overcommits to optimize the limited memory available in 2001. It included QOS controls such as CPU and memory reservations and limits as well as the ability for administrators to split different groups of VMs into different “pools” of resources and within the pool prioritize some VMs over others. It enabled limits on network usage for both outbound and inbound traffic, and ways to cluster multiple data stores to evenly balance storage traffic and usage.

More significantly, VMware treated each ESX host as a standalone server. The obvious next step was to develop a centralized management platform that could coordinate multiple separate servers. This agenda became an urgent need when, in February 2003, Microsoft purchased Connectix and its Virtual Server product. The Redmond giant had a fearsome reputation for crushing competitors from Lotus to WordPerfect to Netscape. VMware had to differentiate itself from Microsoft’s Virtual Server, and do it quickly.

VMware faced two large limitations of the times. First, the rise of virtual appliances was still in the future. Customers would not tolerate a VM appliance running on the same cluster that it managed. Second, Linux had not yet won the Unix wars and a Linux-based management platform was not likely to be widely accepted.

The fastest way for VMware to gain acceptance for its management platform was to develop it  as a Windows application running on bare metal servers. Customers were comfortable with this architecture. VMware further encouraged acceptance by using a standard database backend with MS Access DB, SQL or Oracle.

On December 3, 2003, VMware introduced its management solution, VirtualServer, to the world. When combined with VMotion, this approach proved good enough to win the datacenter wars. Despite the threat from Microsoft and all of the industry doomsayers proclaiming that VMware’s only hope was to slash prices, the virtualization pioneer went on to dominate the industry it created.

NUTANIX – BORN IN THE CLOUD ERA

When Nutanix went to market in 2012, network connectivity was 10Gb, Intel processors came with 8 cores running at 2.7 GHz. Memory was sold in 4GB DIMMs running at 1,600 MHz for $40 a pair. Flash was on its way to becoming ubiquitous. This exponential advancement in resources enabled the software-defined hyperconverged infrastructure (HCI) category. It also meant that Nutanix could automate many of the virtualization functions that administrators configure themselves in a VMware environment.

A Forbes 10/26/2013 article – Will Dheeraj Pandey’s Nutanix Be the Next VMware of the Data Center? reads, “[Nutanix] incorporates the same advanced, distributed software architecture that powers leading public cloud providers, such as Google, Facebook and Amazon — but is tailored for mainstream enterprises and public cloud providers.” Nutanix brought this same web-scale architecture to virtualization by customizing the KVM hypervisor to integrate with, and capitalize upon, the software-defined HCI.

Nutanix virtualization includes integrated management as part of every node; management seamlessly scales out with the environment. And resiliency is enabled using the same distributed web-scale principles utilized by Nutanix for its operating system as well as by the leading cloud providers.

During VMware’s initial growth stage, many smaller shops were still building white box servers, and tower servers represented a decent chunk of the market. This led to a massive number of servers along with all of the storage and networking components supported on VMware’s HCL (Hardware Compatibility List). The need for such widespread legacy compatibility adds complexity. It also requires significant time for virtualization administrators to simply validate interoperability.

In contrast, Nutanix AHV strips the complexity out of the KVM open source hypervisor while also enhancing performance and enterprise capabilities. AHV hardens KVM to meet enterprise security requirements and seamlessly integrates it into Nutanix’s Prism management.

Add a Nutanix node to the cluster, and you’ve automatically extended the virtualization which is managed from the same Prism management plane. Should a node fail, the Acropolis Distributed Storage Fabric (ADSF) automatically self-heals for availability. Impacted VMs are restarted on other nodes.

VMware requires many vCenters in order to manage VMs in multiple locations as well as for redundancy. Nutanix Prism, on the other hand, is a single management control plane that is simple, highly available, and configured optimally with no need for architecting deployment, database or HA. AHV liberates virtualization from the domain of specialists – making it both simple and easily managed by anyone from DevOps teams to DBAs.

CLOUD STACK VS. PROPRIETARY

An idea of a cloud stack is somewhat arbitrary, and certainly ever evolving. Today, some typical components include Linux, Containers (i.e. Kubernetes/Docker), Hadoop, YARN, MySQL, Docker, Spark and so on. Another critical component, of course, is the hypervisor. KVM is a virtualization infrastructure for the Linux Kernel. Linux/KVM is open source, meaning that it has a vast global developer community working to improve the hypervisor. VMware, on the other hand, only has a small fraction of its 22,000 employees working on improving ESX.

KVM is increasingly utilized by cloud providers including four of the leading public cloud providers: AWS, Google Cloud Platform, IBM and Oracle. Public cloud customers care about CPU speeds, storage capacity and IOPS. They don’t care about what hypervisor is running “under the hood”, and they certainly don’t want to pay more for it.

Virtualization as a proprietary stand-alone product has had an incredible run, transforming data centers the world over into hosting environments for virtual machines. But the public cloud era has arguably evolved virtualization from product to feature. As I predicted back in a 2015 blog post, Gartner has now even retired its Magic Quadrant for virtualization.

OPERATIONAL COMPARISONS OF VSPHERE AND NUTANIX AHV

While vCenter today can be run as a Linux-based virtual appliance using VMware Photon OS, it remains a repackaging of the original Windows-based software. Some tasks are simpler, but vCenter still entails operational complexity, does not scale well, is not natively resilient, and is not multi-site. This contrasts with the simplicity resulting from building in management as part of the OS. Consider, for example, the install and cluster setup process for vSphere vs. AHV:

vSphere

  • Install ESXi manually on multiple hosts (assuming no Autodeploy server, but this is rarely seen in the wild).
  • Deploy the vCenter appliance (taking into account complex considerations such as whether to use an internal or external platform service controller)
  • Configure vCenter for high availability (doubling the CPU, memory, storage and licensing required)
  • Individually, and manually, join each ESXi host to vCenter.
  • Form a cluster
  • Enable HA and DRS
  • Create and migrate the hosts from the simple vSwitch to a Distributed vSwitch that keeps the VLANs and port groups in place (If the vCenter that runs the Distributed vSwitch is on the cluster you’re migrating, you’ve got to be extra careful not to knock it offline and create a chicken and egg scenario.
  • If following VMware best practices, create another management port and VLAN for vMotion traffic (potentially multiple management ports if using multi-NIC vMotion).
  • Apply a host profile to ensure all the settings are identical on every host (typically requires putting the hosts into maintenance mode and evacuating all the VMs to apply).
  • Add vROps for capacity planning and right sizing

AHV (NX series)

  • Automated node installation through Nutanix Foundation (either bare metal or discovery)       
  • Create a new cluster or add to an existing cluster (adding new nodes and expanding the cluster is completely automated)
  • Optionally deploy Prism Central (via simple wizard. This enables vROps, vRA and partial NSX replacement).
  • Choose between a single server and scale out (which you can change later).

Even for an expert level VMware administrator, it typically takes around four hours to build a cluster from scratch. This setup time increases by around 10 – 30 minutes per host in the cluster. AHV requires about 45 – 90 minutes depending upon the cluster setup process used, and that time remains constant regardless of the number of hosts.

WHICH VIRTUALIZATION PLATFORM SHOULD AN ORGANIZATION USE?

This question puts the cart before the horse. An evaluation should start with business objectives/outcomes in order to provide context for the decision-making. The next step is to identify requirements. Only at this point does it make sense to evaluate which virtualization platform is more appropriate.

Why vSphere?

VMware is a phenomenally successful and mature company with over 600,000 customers, a massive partner ecosystem and ten thousand channel partners. And, of course, its parent company is the $79B (annual revenues) Dell. Despite the virtualization sea change initiated by public cloud, VMware will be around for a long time.

AHV, as a feature of the Nutanix Enterprise Cloud OS, is only available for Nutanix customers. Organizations not intending to purchase Nutanix should stick with vSphere or another hypervisor. And, if an organization plans to keep 3-tier infrastructure (centralized storage, storage network, servers), then vSphere is optimal because of its 17 years of hooks into legacy storage. But organizations with this inclination may wish to compare 3-tier vs. HCI via a TCO analysis. Even VMware’s CEO, Pat Gelsinger, publicly stated at last year’s VMworld, “All infrastructure will become hyperconverged.”

If an organization has invested heavily in, and uses all the capabilities of, VMware NSX, then vSphere is the only hypervisor supported. If it has older hardware with limited memory, no other hypervisor (today) is as efficient at memory overcommit as vSphere. If a company’s operational policies are tightly woven around vSphere, perhaps including scripting, then it may be cost prohibitive to move away from the hypervisor despite lower licensing, support and operational costs.

If an organization is happy with a primarily on-premises environment and its staff is determined to stick with what they know, then it might not make sense to move off the hypervisor despite lower licensing and operational costs. Similarly, if widely used key applications are only supported on vSphere, then it may not be worth the battle to run them on AHV. After all, VMware runs great on Nutanix. Even when AHV isn’t utilized, vSphere customers benefit from all of the other Nutanix goodness.

What About VMware on AWS?

Why would an organization ever pay more money to run VMware vSphere on AWS? I can see three reasons:

1)  IT wants to stick with 3-tier infrastructure on-premises, making the AWS infrastructure scalability particularly attractive even at a higher cost.

2)  The company wants to quickly expand or contract infrastructure utilizing AWS’s massive capabilities.

3) I strongly suspect the most significant reason: The CIO wants to tell the CEO or Board that she is in “Cloud” without having to significantly change her operational procedures, governance, etc. In other words, VMware on AWS allows the CIO to check the “We’re on Cloud” box though this architecture doesn’t capitalize on many of the true public cloud advantages.

Why Nutanix?

If an organization believes its future is hybrid cloud (and especially the multi-cloud subset), then it may make sense to look at the virtualization platform most aligned with those of public clouds, Nutanix AHV. Customers benefit from a much simpler virtualization platform, lower costs, easier management and better scalability. And Nutanix makes it easy to migrate vSphere and Hyper-V workloads to AHV.

In July 2018, IDC published a study, Cloud Repatriation Accelerates in a Multi-Cloud World, based upon a survey of 400 cloud customers. IDC found that 80% of the participants had repatriated some applications back on-premises. The research organization projected that 50% of all applications will be repatriated over the next two years. Nutanix makes it easy to migrate workloads from AWS to on-premises Nutanix AHV.

NUTANIX MAY BE A COPYCAT – BUT NOT OF VMWARE

Two months ago, in the CRN article quoted at the beginning of this post, VMware COO Sanjay Poonen called Nutanix a “copycat” and a “VMware wannabe”. Whenever a new technology paradigm arises, the old guard will inevitably refer to upstarts as mere copycats. Nonetheless, his statement is partly accurate – Nutanix’s proprietary technology is similar to other great ideas about delivering cloud services, just not ideas from VMware.

Nutanix has openly documented its architecture in the freely available Nutanix Bible, and is committed to open-source software, actively using and contributing code within a variety of communities. Open-source, software-defined everything, and simplicity are core Nutanix principles. This translates not only into a cloud-like approach to virtualization, but also into solutions such as Nutanix Frame which provides a compelling alternative to Amazon Workspaces.

Nutanix diverges from AWS by merging the best of on-premises and public cloud into a multi-cloud platform. Its Enterprise Cloud OS combines agility, service-based architecture, high availability, and continuous 1-click innovation with enterprise operations and consumer-grade design. Customers gain more control and customization capability. And they are able to slash the cost of IT in the process.

 

Huge thanks to Wayne Conrad for conceiving of the focal idea behind this post that each firm’s approach stems from its heritage, as well as for much of the content. Thanks also to Laurent Sandrolini, Mike Wronski, Jason Burroughs, Michael Webster and Wendy Pfeiffer for content and editing.

Disclaimer: The views expressed in this blog are those of the author and not necessarily those of Nutanix, Inc. or any of its other employees or affiliates.

See Also:

Gottcha Pricing from the Cloud Pushes Workloads Back On-Premises. 11/16/2018. Andy Patrizio. NetworkWorld.
Linux Now Dominates Azure. 09/27/2018. Steven J Vaughn-Nichols. ZDNet.
VMware Cloud CTO: Launching VMware on AWS Has Been a Heavy Lift. 07/12/2018. Data Center Knowledge.
The Rise of Native Virtualization. 04/25/2018. Steve Kaplan. By The Bell.
Server Virtualization Software comparison ─ Microsoft Hyper-V vs VMware vSphere vs Citrix XenServer vs Red Hat KVM. 01/25/2018. Jyotsana. ZNET Live.
Improving KVM x86 Nested Virtualization. 12/15/2017. Liron Alon. Linux Foundation.
NF102: Nutanix AHV Basics. 11/15/2017. Mike Wronski. Nutanix.NEXT.
A Deep Dive Into the Benefits of KVM and Open Virtualization. August 2012. David Hsu. IBM.
VMotion, The Story and Confessions. 11/11/2010. Duncan Epping. Yellow Bricks.

This entry was posted in Virtualization and tagged , , , , . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

*