Now that 2018 is coming to a close, it is interesting to look back and see that lots of trends in data center technology are becoming more mainstream. Cloud adoption is ramping up, Kubernetes is bringing containers to the enterprise, and hybrid cloud looks like the blueprint many organizations will be adopting for the future. With some of the recent buzz also around not so “sci-fi” topics like serverless compute and AI/ML, many people have started wondering what the next big trend is going to be.
While I’m not here to make any predictions on what unforeseen breakthrough will be generate all the buzz in 2019, I do know that for many organizations outside of the largest enterprises, hybrid cloud is still a goal that is attainable in the present. Because of that, a lot of organizations that typically deal with traditional data center infrastructure are taking a long, hard look at HCI to help with the on-prem side of the hybrid cloud.
As a Data Center Architect for a VAR, I am constantly bombarded with marketing materials for all the latest products from both big name vendors and smaller niche players. This post is part to inform, and part to selfishly help myself differentiate all of the major HCI players out there today. This is no way a ranking or review of the different products, rather I want to give an unbiased overview of the products that are on the market today.
Built upon the successful UCS infrastructure, HyperFlex is Cisco’s entry into the HCI market. Unlike most of the other HCI products, HyperFlex nodes consist of a rack mountable server as a single node. A ‘typical’ HCI chassis consists of multiple nodes in one box, which means HyperFlex leaves a similar data center footprint as traditional compute from a size, power and cooling perspective. HyperFlex originally included support for VMware, but recently expanded to Hyper-V as well. The size of the nodes does allow for some of the highest available capacity on the market, with high performance from all-flash clusters to blazing fast performance from all NVMe nodes. HyperFlex can also be configured as a stretched cluster, allowing for active-active data centers in different physical locations. Due to HyperFlex being the HCI extension of UCS, it can also integrate and be deployed with InterSight, Cisco’s cloud-based management and analytics platform. With Cisco’s “ACI Anywhere” vision (link provided to a Tech Field Day 17 presentation I attended at Cisco), customers who have deployed ACI certainly have a compelling case to use HyperFlex as a their HCI platform. Cisco also recently partnered with Veeam to deliver an integrated solution that runs on HyperFlex, providing scalability and performance on top of Veeam’s data availability platform.
NetApp HCI is a relatively new player in the field. Built on the SolidFire Element OS platform, there is a lot of talk about whether NetApp HCI is truly ‘HCI’ or not. Each NetApp HCI node is either compute or storage, rather than a convergence of both. That means that a bare minimum of 2 compute nodes (think vSphere HA cluster) and 4 storage nodes (think SolidFire storage cluster resiliency) are needed out of the gate. The upside here is that this is the only solution that can truly scale by either only compute OR storage, rather than both at the same time (or in some instances storage only). It is also the only solution that does not have the HCI tax, which means a storage controller VM does not live within the compute cluster, therefore stealing compute resources away from application data. VMware is currently the only supported hypervisor, but Element OS provides guaranteed storage QoS via IOPS throttling. Since this HCI solution lives within the NetApp ecosystem, that also means NAS storage via ONTAP file services and full integration with the NetApp Data Fabric.
Nutanix has made a big splash independent of the major enterprise vendors. It has been one of the leaders in HCI for a while now and certainly generates a lot of buzz. The biggest differentiator for Nutanix may be that they bring their own hypervisor to the table. While VMware or Hyper-V is supported on Nutanix, the option to run their in-house AHV hypervisor is also available. The HCI environment itself is accessed via Nutanix’s Acropolis management platform, and AHV is integrated directly into Acropolis at no extra cost. That means a single plane of glass for management of the hardware platform and the hypervisor itself, with no extra per-core licensing or cluster management software necessary. On top of that, Nutanix provides many other benefits, including metro availability, it’s own NAS storage platform, and backup and DR capabilities. Nutanix has a strong focus on public cloud integration, first with Cloud Connect, which provided the ability to back up VMs to AWS or Azure. Later, Xi Cloud Services was released, which delivered a broader set of integrated cloud DR services, in addition to cloud optimization, governance, DaaS and IoT. Nutanix also built in-house solutions for application automation (Calm), database services (Era) and Kubernetes distribution (Karbon). While Nutanix fulfills the on-paper requirements for HCI, they are very focused on delivering a breadth of services within the Acropolis platform itself.
One of the HCI products that made a big splash early on is SimpliVity, which was later acquired by HPE. SimpliVity follows the traditional HCI form factor, but also includes the OmniStack Accelerator card, which is used to offload data deduplication and compression, as well as flash acceleration. That means that even though the controller VMs share hypervisor compute resources, a good chunk of the data platform is offloaded directly to the OAC. This also helps with SimpliVity being able to perform what they call global federation, which means that data is deduped globally across all SimpliVity nodes in multiple locations. That greatly improves SimpliVity’s ability to replicate and protect data across all devices, as only the bits that don’t already exist on other nodes need to be transferred across the WAN. SimpliVity also provides a plugin for vCenter that allows for management within the vSphere client, including the ability to perform per-VM backups and restores, clones, create backup policies, view performance, etc. These actions are performed natively on SimpliVity, but utilize the vSphere client as the single pane of glass from a management perspective.
Dell EMC VxRail
Dell EMC’s entry into the field is VxRail, which is a jointly engineered solution that focuses completely on the VMware ecosystem. The hypervisor (obviously) is ESXi and the underlying storage platform is vSAN. VxRail essentially brings the “VMware shop” into the market as a HCI appliance. Everything you would expect from the VMware landscape comes fully integrated and pre-configured in the HCI form factor. On top of that, VMware brings automated lifecycle management, as is becoming standard in the latest SDDC offerings from VMware. There is also the option to include NVMe cache drives and enhanced services if Dell EMC network switches are used for uplinks. The big kicker here is the fast track to having full on-prem integration with VMware’s cloud services, including VMware on AWS. VMware has made it clear that their vision is to be the hybrid cloud solution of the future, and VxRail simplifies the infrastructure footprint that is needed to best integrate with all flavors of VMware. As those offerings mature, the automated lifecycle management will ensure that the on-prem side of the hybrid cloud is always lock-step with the rapid advancement happening in VMware cloud.
HCI’s growth in the market has been steadily increasing, with estimates predicting a ten-fold increase by 2023. I’ve certainly seen this in the field, as HCI is almost becoming a requirement that customers expect to look at when it comes time to refresh traditional converged infrastructure solutions. At the end of the day, all of the products mentioned above help decrease the size of a physical data center footprint and are more than capable of running traditional workloads. From a performance and features standpoint, they will more than likely be at least on par with, if not much more so than an older generation of traditional hardware. Many customers already have a vendor preference, so it may be easy to choose Simplivity solely because ProLiant servers and 3Par storage needs refreshed, and there is still a great vendor relationship with HPE. It is really the fringe use cases that will more than likely drive a customer towards one specific product over another. Are you currently a UCS customer and want to add on to your existing infrastructure? Do you have an application that relies on highly performing databases with large amounts of IO? Are you planning on providing multi-tenancy that will require different levels of performance? Are you looking to easily integrate with multiple public cloud providers or VMC on AWS? The answers to these types of questions will more than likely narrow the road…and for at least the next few years there is a great chance it will be one of these HCI options waiting at the end of that road.