Do you understand hyperconvergence?
Everyone needs more storage — but what should be your next purchase? Hyperconvergence has emerged as the buzzword to rule them all. Offering to solve storage needs simply, breaking down silos and offering straightforward scaling. Is this, however, true? Are all hyperconverged systems created equal?
Hyperconvergence, or more accurately hyperconverged infrastructure (HCI), is a little confusing. This isn’t so much because of what it is. Hyperconvergence, in a nutshell, is the collapsing of distinct hardware provisions for CPU, networking, memory and storage into virtualised aspects of a single and interchangeable piece of kit. The problem is that the vendors aren’t consistent in how they talk about these things. Then there is hyperconverged architecture, ‘converged infrastructure’ CI and, oh yeah, HCIS — hyperconverged integrated systems.
We are going to explain all of this, and we are going to do so while investigating some of the specifics of two big players in the ‘hyperconverged’ space: Nutanix and NetApp.
Nutanix: A Hyperconverged Leader Ruling in the Cloud
Nutanix burst into the hyperconverged scene in 2009. By 2013, the company reached a valuation of $1 billion. They have continued to evolve and are now back at the top of Gartner’s ‘Magic Quadrant’ 2018 report on hyperconverged vendors.
Designed from the ground up for virtualised workloads, Nutanix is hyperconverged to its core.* This means that they sell virtually partitioned pieces of hardware that perform all compute and storage capabilities in a single package. Their hardware can be deployed on-site or in a data centre.
They do offer a ‘bare-metal’ configuration called Acropolis Block Services (ABS) for applications that cannot be supported by a virtualised environment.* Although ABS could be purchased on its own, it is primarily sold as a way to incrementally grow into a fully virtualised and hyperconverged environment. ABS integrates with the Nutanix software solution Prism that drives their entire ecosystem.
Prism follows in the footsteps of leading storage management software tools. It delivers streamlined workflows, provisioning, compression, deduplication and redundancy settings on a broad and granular level. It even uses machine learning technology to execute active troubleshooting and automate common tasks. Prism provides actionable insights to optimise performance through an easy to use HTML5 interface.
Nutanix hardware can be used for many purposes. Where Nutanix focuses their efforts, however, is creating hybrid cloud environments that ‘simplify management’ and allow for non-disruptive ‘one-click’ scaling. HCI is already virtualised, so interfacing with the cloud is easy. Nutanix allows you to build a private cloud using their HCI hardware, interface with their public cloud offering (Nutanix Enterprise Cloud OS) and your own on-site Nutanix hardware using a single point of control.
Although it offers a powerful, seamless and simple option for internal use and deployment, Nutanix is a broadly closed ecosystem. However, it can be natively integrated with several other virtualised solutions including VMware, Microsoft HyperV and Citrix.* All common tasks including deploying, cloning and protecting VMs can be managed centrally through Prism.
NetApp: A Traditional Specialist Entering New Spaces
NetApp is a long-time dominant force in enterprise storage. Founded in the early 1990s, NetApp led the way in NAS technology. They are now the second largest vendor in external enterprise storage, pulling in $890 million in 1Q18 revenue.* This September, NetApp announced a strategic partnership with Lenovo that will expand their in-house offerings, allowing them to compete directly with the giant IT conglomerates like Dell and HPE in every market. Although, they have gone toe-to-toe with these behemoths for years, specialising in third-party integrations and outpacing HPE in enterprise storage.
But, NetApp is new to hyperconvergence. In 2015, they acquired SolidFire, bringing all-flash arrays into their line-up. SolidFire hardware effectively forms the basis of the NetApp HCI solution, launched in 2017. NetApp HCI consists of industry standard high-density, four server and two rack units.* But, rather than using a hypervisor to run its storage management processes (as done by Nutanix), NetApp dedicates particular nodes to act as servers — running its SolidFire Element OS.
This design delivers simple hardware building blocks for an easy, dynamic and scale-out storage experience — exactly what HCI is designed to do. But, the use of designated and separate nodes to act as servers has led some to call NetApp HCI a converged infrastructure (CI) solution, or even, ‘disaggregated software-defined architecture’.* The truth, however, is this delivers an outcome nearly identical to that of ‘true’ HCI. It just gets there in a slightly different way. In fact, the NetApp approach might bring some advantages, and market trends seem to be completely blurring these lines. But, for that, we need to back up.
HCI, CI, HCIS — What Do These Terms Mean and Do We Need Hyperconvergence Anyway?
Traditionally, the data centre is comprised of servers, networks and storage components. These are specialised pieces of hardware that can be bought separately. Theoretically, this enables maximum flexibility. However, it creates compatibility challenges.
Since the late 2000s, storage vendors began selling ‘pre-integrated’ boxes that converged network storage and servers into a single, off-the-shelf solution. These came to be known as converged infrastructure (CI). HCI (hyperconvergence) smashes all of these ‘nodes’ into one piece of hardware. Rather than being multiple pieces of kit packaged together, HCI offers them in one piece that is virtually subdivided. This was a move spurred by cost. It is just cheaper to build hardware this way, rather than to construct it out of many parts.
HCI hardware is also smaller and uses less energy. This means that your upfront costs will be higher than traditional hardware. However, maintenance and operational costs will be less. But, a portion of the server’s CPU and RAM are used to manage the system, slightly diminishing its capacity compared to CI in like-for-like testing. The cost savings, however, in setup and deployment generally mitigate this issue.
The thing, though, is that not all HCI units are exactly the same. You can’t just manipulate the server and storage portions of your HCI box, they come pre-set. Early HCI vendors developed pre-scripted configuration (with different proportions of storage and compute capabilities) to cater to different types of operational needs. But, this has always created the issue of over-provisioning while scaling.
The irony is that although hyperconvergence (and converged infrastructure) was developed to simplify scaling, many systems cannot do so flexible. Expensive HCI blocks that either overprovision compute or storage have to be bought when only one is needed. To get around this, HCI vendors started selling compute only and storage only nodes. Although Nutanix resisted this move, they now support storage only scaling.*
The thing is that we are now back to what NetApp has to offer. If you build a Nutanix array of server/storage HCI nodes, and then add on some Nutanix storage only nodes, are we really in a different place than a NetApp solution that has used separate server and storage nodes from the beginning?
Even if the answer to that question is yes, it doesn’t seem meaningful. What does seem meaningful is that the NetApp solution will also allow you to scale your server node independently, delivering even more flexibility. This hardware segregation also allows NetApp to make more sound guarantees regarding the latency of its Element OS control system.* The downside of this partitioning is that you have to buy four nodes, otherwise the system will not work. There is a minimum buy-in, but you get more flexible scalability.
NetApp vs. Nutanix: Hyperconverged or Not, Maybe It Doesn’t Matter
Gartner uses the terms HCI (hyperconverged infrastructure) and HCIS — hyperconverged integrated systems.* Although HCI describes something very specific — virtually partitioned computer servers and storage — HCIS is defined by an outcome orientation perspective. HCIS describes systems that might use HCI hardware, but broadly focus on software-defined solutions that make storage simpler. Although Gartner does not consider NetApp to sell an HCI solution, they named them an ‘Evolutionary Disrupter’ in the HCIS market in 2017.*
The reality is that hyperconverged vendors are breaking apart their integrated blocks and selling even more flexible solutions that, ultimately, look a little bit more like where we started before — converged infrastructure. The hardware is just better, faster, integrated with the cloud and controlled by more advanced software solutions. When looking at outcomes, HCI might not be the best term anymore.
Fundamentally, although both NetApp and Nutanix come at the hardware from a different perspective, they both sell very comparable products that compete for customers. From that perspective, HCIS makes more sense as a means of describing the nature of the market.
Summary: Nutanix vs. NetApp — Flexibility vs. Out-of-the-Box Functionality
The best decision for you might have to do more with what you already have. Although hyperconvergence is about easy scaling and breaking down silos, hyperconverged storage segments can end up as their own silos. They are easy to scale and easy to integrate, but only when it comes to that hyperconverged system. Integrating new hyperconverged hardware into an existing network is just as hard as integrating any new piece of hardware.
NetApp HCI runs on Element OS, from their SolidFire range. This is a quality operating system that provides all of the simplicity and functionality you need. If you have traditional SolidFire all-flash arrays, you will be able to unite all of that storage in one system.
Beyond the fact that NetApp sells a wide range of enterprise storage products, they are notoriously good at integrating with third-party options. This comes from ONTAP, their other software solution. Using ‘FlexArray Virtualisation’, ONTAP can basically integrate any third-party hardware — bringing NetApp HCI into the mix is no problem. If you are operating in a legacy environment made up of many different vendors, this is very useful.
ONTAP also comes with troubleshooting, replication, duplication and provisioning features that are somewhat unique and industry-leading. Beyond that, it offers their patented WAFL (Write Anywhere File Layout) system. This is only really useful for hybrid arrays, allowing SSD to operate much more closely to the speed of flash, and will have no impact on your hyperconverged system. But, the benefits of NetApp really come from looking at the wider picture, and your desire to build a larger ‘ad hoc’ storage system from a lot of existing pieces. This flexibility is mirrored in the scaling capabilities of the NetApp HCI solution, which is able to add storage or compute independently in modular capacities.
On the other hand, Nutanix excels at specifically what it is designed to do — delivering a powerful and seamless hyperconverged and cloud experience within their own ecosystem. The Nutanix software system Prism has machine learning capabilities that are not matched by NetApp, and the whole system was built for virtualised storage from the ground up. It is a newer, more advanced option. There is a reason industry-leading businesses turn to Nutanix for their hyperconverged and private cloud needs — they deliver a fantastic product.
The problem is that Nutanix is harder to integrate into a legacy environment, and does not offer the same type of support for non-hyperconverged components. If what you want is a powerful, cloud-focused and truly virtualised solution — you want to buy Nutanix. You will get a machine learning powered operating system on top. The danger is that the whole thing could end up as a silo that will have to be managed separately from your other ecosystems and that occasionally you will have to overprovision storage to upgrade server capabilities.
The flexibility of NetApp ensures that you will never have to overprovision. But, where NetApp shines is from the wider data centre perspective. They have more to offer than just hyperconvergence, and, as we have covered, they don’t really have an HCI solution. If you are looking to unify your data centre using a single software storage solution, NetApp is a great place to start. Although you might be able to build something great, you should be aware that it will take a little more effort. For out-of-the-box simplicity and power, Nutanix is your solution.
*IT Central Station: NetApp vs Nutanix
*Nutanix Acropolis Block Services
*Worldwide Enterprise Storage Market Grew 34.4% during the First Quarter of 2018, According to IDC
* NetApp HCI: Kind of, sort of, but not really hyper-converged
*NetApp HCI: More converged than hyperconverged?
*Nutanix Scalability – Part 1 – Storage Capacity
*Magic Quadrant for Hyperconverged Infrastructure
*Competitive Landscape: Hyperconverged Integrated Systems