Your Resource for All Things Apps, Ops, and Infrastructure

In-Depth Product Review: Nutanix’ “Invisible Infrastructure”

Nutanix has been available in the market since 2011.  From 2009 through the end of 2014, most would consider Nutanix (as well as most hyperconverged infrastructure technology) to be at the beginning of the bell curve of industry adoption. Their technology provides significant advantages to traditional as well as converged data centers at a similar price point. However, as a still new and proprietary technology, most large enterprises have stayed away from it because data center managers have been unwilling to accept an unknown amount of risk.

A combination of factors that include demand for management simplicity, natural “fit” with web 3.0 applications, low entry point, and theoretically infinite expandability have enabled Nutanix to break through into large enterprise. We expect 2016 to be a stellar year for Nutanix.

As a specialist technical architect at AHEAD, my primary responsibility is to be a subject manager expert on converged (CI) and hyperconverged (HCI) infrastructure. As I review and evaluate the different products in the marketplace, I will be writing a series of blogs and will be focusing on a single product per post. I started with a post that covered Hitachi UCP, and for this second post, I’ll cover Nutanix’ “invisible infrastructure” which falls in the HCI family.

Nutanix, founded in 2009.Company Intro and CI History

Nutanix was started in 2009 in San Jose, California and shipped its first product in 2011. The founders of Nutanix studied what companies like Google and Facebook were doing in their data centers, which had been powered by inexpensive commodity server hardware. Rather than investing in expensive high availability designs with redundant hardware, they figured out that operational resiliency, data redundancy, and a seamless customer experience could be provided through engineering of intelligent software. If software could be engineered with data services such that it could manage commodity hardware, they would be able to introduce a new paradigm for general data center use. The question was, how could they take the single-purpose Google or Facebook architectures and develop a product suitable for the traditional applications found in most corporate environments.

Since several hypervisors already existed in the marketplace that manage workloads from a processor, memory, and network perspective, Nutanix focused specifically on development of a solution to replace the traditional storage array. Nutanix developed software that stitches local storage together from several independent servers to make it look like a traditional pool of shared storage similar to what could be presented by storage area network to a hypervisor based virtualization platform. By doing this, the major hypervisors could be installed on Nutanix nodes. They could also gain some of the same availability benefits and probably higher performance than could be typically provided by traditional storage arrays, but without the complexity of storage arrays or storage area networks. Since the initial launch in 2009, they have received $300M+ in funding, which places them at a valuation of over $2 billion.

Today Nutanix’ product line can be divided into two software components (Prism and Acropolis) along with two options for hardware platforms. Nutanix ships on either Nutanix branded nodes (manufactured by SuperMicro) or on Dell rack servers. In both cases, the systems are powered by Nutanix software. Prism is Nutanix’ management platform that has been available since product inception, and Acropolis is Nutanix’ recently announced hypervisor that allows Nutanix to provide a KVM based solution competitive to Hyper-V and VMware vSphere.

For the purposes of this post I will omit the Dell hardware models, just note that they are comparable in performance, functionality, and price to the Nutanix branded hardware described below.

Architecture & Features

  • Product Range
    • A Nutanix node contains CPU, RAM, and local disk along with two to four network ports. The hardware models, ranging from 1000 to 9000 contain different mixes of these resources which allow customers to obtain the nodes that best fit their workload requirements.
    • Model options include graphics acceleration powered by NVidia for VDI, all-flash storage for high IO random workloads, and capacity nodes for low IO, archival workloads.
    • Nutanix offers three different software versions, with increasing number of features as you move from starter through advanced to enterprise. These can be paired up to the requirements of the workload.
  • Scalability Interdependence
    • Nutanix, like most other hyper-converged players, was initially heavily criticized by competitors as having no scalability interdependence. The argument was that if a workload needed more storage but no more compute resources, solutions like Nutanix would scale inefficiently because an additional node would have to be purchased with additional compute resources.
    • In reality, Nutanix provides so many hardware options that it arguably scales as efficiently as traditional infrastructure. This is the case even more now that Nutanix has released storage-only nodes that can run Acropolis (Nutanix’s free hypervisor), meaning that if you are a VMware or Hyper-V environment, you don’t have to pay for the hypervisor OS.
  • Flexibility
    • Nutanix is very flexible in many regards; to start it officially supports Hyper-V, VMware, and KVM (Acropolis).
    • The Prism management software can manage multiple clusters and each cluster can run different hypervisors.
    • Nutanix hardware nodes can be customized to have different Intel Xeon processors, different RAM configuration, different NIC speeds and types, and different disk sizes.
    • Nutanix clusters require having a minimum of three nodes, and for most node types you can mix and match nodes types in a cluster. This is a nice feature in theory, but in practice it doesn’t always make sense to mix node types in a cluster. Again, it all depends on the workload characteristics.

Data Protection, Replication, High Availability

  • Data protection
    • Nutanix supports VM-level snapshots which can be maintained on the local Nutanix cluster or replicated to another local or remote cluster. These snapshots can be manually mounted to other VMs and made accessible. This is by no means an enterprise-level backup solution but, like all other aspects of Nutanix, is fully scriptable via REST.
  • Replication Solutions
    • Nutanix includes local and remote replication between Nutanix clusters which is powered by storage snapshots. These can be on-demand or scheduled.
  • Highly Available Solutions
    • The enterprise software package also unlocks metro clusters, otherwise known as synchronous replication (assuming that distance and/or latency limitations aren’t exceeded).

Primary Components

  • Compute
    • Compute resources are provided by Nutanix nodes come in a form factor of 1, 2 or 4 per Nutanix Chassis. The Nutanix chassis only shares power across nodes unlike other chassis technology that also shares network fabric.
    • Nutanix hardware is currently powered by Intel Xeon processors with the largest processor option being the E5-2699v3 which has 16 cores. The largest node configuration has two sockets and 24 DIMM slots, supporting up to 768 GB of RAM.
    • The compute options are in-line with standard-issue enterprise servers which can run all but the most demanding workloads, such as in-memory database servers which typically have E7 Xeon processors and 1 TB to 6 TB of RAM.
    • Certain Nutanix models support 1 or 2 NVidia K1 and K2 cards which are typically used for VDI workloads that require graphics acceleration.
  • Storage Array
    • This is the section where HCI products shine. There is no dedicated storage array, no Fiber Channel or iSCSI SAN, no (public) CIFS/NFS.  Instead, storage is clustered across nodes in a redundant manner and presented back to each node, in this case via NFS. Nutanix provides fantastic storage performance through a combination of caching hot data to RAM and then SSD as well as through data locality, where VMs running on nodes have their data on local drives, avoiding having to read/write through the 10gb network.
    • This “storage magic” is made possible by the CVM, a virtual machine running on each Nutanix node. Local disks on each node are passed directly to the CVM which creates a cluster with other CVMs on other Nutanix nodes. Note that RAID is not used, so there are no write penalties. Instead Nutanix makes multiple copies of data, across nodes to provide data redundancy.
    • Nutanix supports inline and post-process deduplication, compression, and erasure coding. This goes a long way towards overcoming their data redundancy scheme where data is copied three or four times to provide protection against a node failure.
    • Nutanix will be releasing a file service in Q4 of 2015 which will effectively provide SMB2 access, replacing the need for virtualized file servers and which will leverage all the storage benefits currently available listed above.
    • Nutanix has built-in intelligence for vSphere storage APIs and is aware of heavy clone operations common to VDI environments, providing improved performance by copying a VDI gold image to each Nutanix node that hosts VDI desktop sessions.
  • Storage Fabric
    • Nutanix leverages standard 10Gb (and soon 40Gb) networking for both production data as well as storage (NFS) access. This gives customers the option to pick the network solution of their choice.
    • Each node has two or four 10GB ports that support common network aggregation protocols that are available to the hypervisors. Collapsing management, vmotion, NFS traffic, and data traffic makes things simple (less NICs to manage) but it also has the potential of scaring away customers that want to physical segregate some of these services.
    • This flexibility is double-edged sword; critics will say that giving customers this option can introduce compatibility issues when compared to CI technology where the network and storage stacks are pre-determined and immutable.
  • Network
    • As mentioned earlier, the network fabric is 10Gb (and soon 40Gb) and any network switches can be used.
    • Compared to some converged infrastructure products, notably those that leverage Cisco UCS, the Nutanix network aggregation tends to be less efficient, because each node consumes two or four network ports. However, since Nutanix does not use Cisco Fabric Interconnects, the overall cost for networking is likely to be close to neutral when compared to Cisco UCS based solutions.

Hypervisor Compatibility

  • Nutanix officially supports Microsoft Hyper-V, VMware vSphere, and Linux KVM. Nutanix recently released their own version of KVM named Acropolis. The primary benefit of this release is that Nutanix provides vCenter-like functionality built into the Prism interface, the same management portal used to manage infrastructure.
  • The Acropolis offering is interesting because it avoids having to pay for VMware, and if a customer doesn’t use any Microsoft, then the customer stops having to pay for any operating system costs. Most enterprise customers aren’t ready to put tier one apps on this relatively new KVM offering, but it could be a good fit for tier two apps and/or Linux-based workloads.
  • Unlike VCE, HP, and a few others, Nutanix is not an official vSphere OEM partner. Nutanix has seen some bad press about this, primarily around the thought that Nutanix is not able/capable of providing support for vSphere. In reality, Nutanix has been able to provide solid support by maintaining a solid team of VMware experts, including several VCDXs.  However, based on the competitive marketplace and, specifically, competition with both vSAN and vSphere (via Acropolis), VMware does not seem interested in partnering with Nutanix or providing Nutanix an OEM status.

Management Features

  • Centralized Visibility
    • Prism is the interface used to manage all things Nutanix. From the web-based portal you have GUI access to all performance, configuration, and alert information for your Nutanix cluster. Prism runs on each CVM and is clustered, making the management solution redundant.
    • Prism Central is a centralized tool that allows you to manage multiple Nutanix clusters from a single interface. It aggregates alerts and performance data and allows you to click into each cluster for more detailed information.
  • Centralized Management
    • Prism gives you access not just to retrieve information but also to make changes to the environment. Through the tool, you can add or remove nodes, configure data protection, backup, and replication settings and more.
    • Prism is 100% REST-based and, in fact, all the commands in the GUI are actually REST calls; Nutanix led with this implementation to guarantee that their management solution was fully scriptable.
    • For KVM installations, all your VM management (like VM creation), power on/off, vmotion, etc is done through the same screen. For vSphere and Hyper-V you would go to your appropriate management tool for that aspect of management.
  • Automation
    • One of Nutanix’s key value propositions is simplicity, and this is achieved partly through the automation capabilities that it provides.
    • Nutanix currently offers single-click updates for firmware and hypervisors. In reality it’s closer to three clicks, but it is fully automated and light-years ahead of most of their competitors. The only thing you need to worry about is rebooting VMs to take care of VM tools and virtual machine levels.
    • Think of how much time and planning is required in upgrading the firmware of a 20 ESX host environment and then upgrading vSphere. Now think of the value of having all that automated.
    • Nutanix is very REST friendly, in fact the Prism interface has a built-in REST library that you can leverage. Nutanix also has cmdlets for Powershell, a java-based CLI called nCLI and the usual option to SSH into nodes for advanced management.
  • Monitoring and Alerting
    • Again, Prism shines here by monitoring the health of your infrastructure. Key data is available the moment that you log in, and additional storage, CPU, and memory details are available by drilling into each graphic.
    • Since Nutanix doesn’t use RAID, it provides additional interesting information, such as how much disk capacity is consumed on a disk-by-disk basis. This is important because it lets you know whether you need to plan for more SSD heavy nodes in the future.
    • Beyond excellent performance metrics, Prism is aware of the hypervisor that you use and version, the number of VMs and their characteristics, and it automatically groups views for you by things like operating system, host, memory, and disk capacity.
  • Pricing
    • When comparing Nutanix to other solutions, it is important to look at total TCO in your evaluation. Total TCO may include hypervisor, network switches, ports, SFPs, cabling, racks, power, etc. that may not show up on an actual product BOM but are real cost considerations to take into account.
    • Once you account for all of the components related to a Nutanix deployment, Nutanix solutions tend to be comparably priced to traditional enterprise-grade systems from manufacturers like HP, Hitachi, Dell, etc.
  • Integration
    • Nutanix does not have plugins for vCenter, which means that administrators will need to have both vCenter and Prism open. However, since both are now web interfaces this isn’t a terrible issue.
    • Nutanix does have plugins for Microsoft SCCVM to enable certain features such as fast clones.

Nutanix GUI.

Market Summary

Nutanix has been in the market since 2011. Through most of their history into 2015, most would consider Nutanix (as well as most HCI technology) to be at the beginning of the bell curve of industry adoption. Nutanix has experienced a number of ecosystem changes which have propelled both Nutanix and hyperconverged infrastructure to market prominence:

  • Gartner recognized Nutanix (and SimpliVity) in the visionary quadrant of the “Integrated Systems” magic quadrant in 2014, before most people were talking about HCI.
  • Both VMware (through EVO:Rail) in 2014, and then VCE (through VxRack) in 2015 have introduced competitive HCI type appliance architectures to compete with a similar approach to HCI deployments.
  • Gartner then recognized Nutanix in 2015 on their new MQ as being not only visionary, but also able to execute, moving them into the Leaders quadrant.
  • Our experience and observations indicate that 2015 is a break-out year for Nutanix as they begin to take orders from many Fortune 1000 clients for specific use cases – moving them quickly from early adopter up the bell curve of adoption.

In Conclusion…

Nutanix is the company that technologists and infrastructure nerds fell in love with years ago due to their unique and inherently powerful design. Until recently, the innovations provided by Nutanix could not overcome the inherent risks of new products seen by corporate infrastructure managers, which meant that Nutanix sales focused primarily in the mid-market and non-production workloads in enterprises.

In the last six months, we have seen a significant shift in favor of Nutanix, with several large organizations making significant investments in the Nutanix platform. This is a sign that enterprises believe that the Nutanix platform is proven for some tier one applications, or at least that the added value of Nutanix overcomes any remaining concerns about the still relatively new platform.

In the AHEAD Lab and Briefing Center, we have a briefing dedicated to exploring the hyperconverged infrastructure market and crafting a strategy. We can also demo the leading converged and hyperconverged technologies, including Nutanix. Learn more about the briefing and get in touch with us below.

Lay Your Path for End-User Computing.


Conduct a Health Check on your ransomware resiliency and work fearlessly toward the future.

Subscribe to the AHEAD i/o Newsletter