BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
After you gain a broad understanding of the options available in the hyper-converged infrastructure market, you should compile a list of your requirements and match those against HCI products.
HCI vendors feature different architectures, features and capabilities. Your applications and workloads also have unique characteristics and needs. Moreover, your company will have different growth needs from other companies, so it is important to consider HCI platforms that can grow with your organization.
Here, we examine the key features and functionality of systems from 10 leading HCI vendors: Cisco, Datrium, Dell EMC, Hewlett Packard Enterprise (HPE), Maxta, NetApp, Nutanix, Pivot3, Scale Computing and VMware.
Key hyper-converged infrastructure components
Cluster size. This is a key factor in your buying decision. HCI products have different capabilities. Some support adding up to eight nodes per cluster, while others support up to 64 nodes. But that doesn't mean an HCI system that supports a smaller number of nodes is bad. Sometimes, that HCI product is perfect for your workloads.
For example, if you are working in a smaller organization or need to deploy hyper-convergence for a discrete workload, you might consider an SMB-centric product. Scale Computing enables you to deploy a cluster with one to eight nodes and possibly more if prearranged with the company prior to deployment. HPE's latest release to its SimpliVity HCI portfolio includes a product that is designed to fit the space constraints of branch offices or SMBs. Nutanix also sells an SMB-centric version of its product.
On the other hand, if you are in a larger organization and you want to move all of your workloads to an HCI environment, VMware vSAN's 64-node support might be a good fit. However, you will eventually hit practical limits that will vary depending on the type of workloads you are running. There may come a point where the chatter among all of the nodes in your cluster could result in performance degradation. But the number of nodes that it takes to get to this point depends on the product architecture and the workload makeup.
Using extensive research into the hyper-converged infrastructure market, TechTarget editors focused this article series on 10 market-leading vendors that offer software-defined storage and appliances. Our research included data from TechTarget surveys and reports from other well-respected research firms, including Gartner.
Other HCI products take a different approach to scaling because of their architectural differences. For example, there are products in our lineup that scale compute and storage linearly. HCI products from Datrium and NetApp provide independent scaling of storage and compute.
The table below provides an overview of current node maximums, which can change over time. Please note that some vendors, including Maxta, do not impose limits on the number of nodes in a cluster. However, the hypervisor itself may impose a limit. For example, VMware vSphere 6.7 supports up to 64 nodes in a single cluster.
Sizing nodes. Another decision you will need to make is connected to cluster size but is more about how you size each individual node. This is sometimes a choice between software and hardware, too. With a software-centric product, you can usually configure your individual nodes however you like -- to a point. Within the guidelines of the hardware compatibility list defined by the HCI software vendor, you can configure each individual node with as much RAM, processor and storage type and capacity as you need. HCI vendors Maxta and VMware vSAN provide highly configurable products.
As you review the other products from HCI vendors, including Cisco HyperFlex, HPE SimpliVity, NetApp, Nutanix and Pivot3, there are primarily hardware options to consider, and you will need to select one of the vendor's standard configurations. Typically, these vendors have various models with a range of RAM and storage capacity, but you are still limited to the vendor's list.
Some hardware-centric vendors, such as Datrium, allow you to purchase just their software and install it on hardware you buy separately, although you can still buy their integrated hardware if you like. It has been reported that Nutanix wants to move in this direction, and given a sufficient deal size, you can expect other vendors to also consider that option.
Storage type. Storage options vary wildly among vendors. You can choose hard drive-based nodes, all-flash nodes with nothing but solid-state storage or hybrid nodes combining hard drives and flash storage. Of course, cost may play a factor here. Hard drives remain less expensive than flash, but flash products provide substantially more performance. The storage choice you make depends on the kind of applications you need to support. If you have few workloads or workloads that do not push storage resources too hard, a hard drive-based or hybrid node may be suitable. Otherwise, you will want to look at all-flash node options.
You may also want the ability to scale storage independently of compute. The software-based options, which include Maxta and VMware vSAN, provide this ability to a point. You can add storage to individual nodes. But, even with these products, you still need to add RAM and compute as you add more storage nodes.
Other HCI products enable you to scale storage a bit more independently. Nutanix, for example, lets you add storage-centric nodes to a cluster. These nodes feature a lot of storage capacity but relatively little compute and RAM. Datrium and NetApp enable you to add storage-only nodes to your cluster. HPE SimpliVity also allows you to customize node types a bit. For each standard node that has both compute and storage, you can add a compute-only node that uses storage from other nodes in the cluster. Pivot3 allows a similar scenario. Because Pivot3 uses iSCSI as the storage protocol, you can connect non-Pivot3 compute nodes to Pivot3 compute and storage nodes. Cisco's HyperFlex uses NFS and provides a similar outcome.
Media type and storage scaling independence are two storage considerations, but do not forget about how the storage is managed. Virtual storage appliance (VSA)-based products, such as those provided by Maxta, HPE SimpliVity, Nutanix and most offerings that provide multi-hypervisor support, offer more flexibility, such as the ability to operate multiple hypervisors. Under a VSA model, the storage is managed by a virtual machine (VM) running in the environment. Alternatively, you can choose an HCI product that skips the VSA in favor of directly accessing the hardware. But that approach generally locks you into a specific hypervisor. However, it can make scaling less of a challenge and can sometimes improve overall performance.
Hypervisor support. Some HCI products support multiple hypervisors, and others support only one. For example, VMware vSAN supports only vSphere. Scale Computing bundles Kernel-based Virtual Machine (KVM) as the heart of its platforms. The table below examines the hypervisors supported by each of the products in this roundup.
But direct hypervisor support is only part of the story. One question you should ask is how the vendor provides its hypervisor support. Take Nutanix, for example. A common concern with hyper-convergence is the need to buy new hypervisor licenses each time you add another node, even if you are only adding a node to support storage expansion needs. Nutanix has its own heavily customized KVM-based hypervisor known as AHV. With AHV, you can buy storage-only nodes running AHV and add them to your Nutanix cluster, even if the other nodes in the cluster are running another hypervisor, such as vSphere. This enables you to expand your cluster's storage capacity without having to buy more vSphere licenses.
Vendor lock-in. This phrase terrifies IT professionals. Some think that, by going with a single vendor, bad things will happen and that the chosen vendor no longer has an incentive to be fair. Prospective HCI customers may have lock-in fears, too, particularly when you consider that they may have to commit to a new server, storage and hypervisor vendor.
Choose wisely. Almost every vendor will allow you to undertake a proof of concept or trial period. Use it to put the product through a complete battery of tests and test your real workloads on it. Find out what the vendor's growth and scaling experience will be like. Call support. Review the company's financials.
And remember lock-in only exists for a single replacement cycle. If necessary, you can move to a new platform after you have fully depreciated what you have. Or, in the worst case, if you need to exit the platform earlier, most transitions are pretty simple because all of the workloads in a hyper-converged environment are virtualized. Well, almost. Some HCI products enable bare-metal application support. Datrium, for example, provides bare-metal services alongside your virtualization workloads.
Multi-cloud support. Organizations across the globe are adopting private cloud options, and hybrid and multi-cloud architectures are now the norm. HCI vendors are listening. Nutanix has tried hard to define the term enterprise cloud as a way to help people understand that the platform is a part of a broader cloud strategy. Cisco's HyperFlex 3.0 release includes ways in which the platform can integrate with public cloud services. NetApp's HCI product can be a full participant in the company's cloud-centric Data Fabric vision. Scale Computing has a partnership with Google to create a hybrid cloud environment. Pivot3 recently introduced Pivot3 Cloud Edition with initial support for AWS. The point here is to stay current with what HCI vendors are doing with cloud. It is rapidly changing, and it is clear that HCI is not just for the private cloud anymore.
Container deployment. Workloads are no longer all based on VMs. Containers are becoming a more common fixture in the data center. Every HCI platform on the market can run containers without any modifications. Containers can run seamlessly inside VMs, which, in turn, run atop your HCI environment. Some HCI vendors have formal programs to support containers. Maxta, for example, recently announced support for RedHat's OpenShift.
Additional benefit comes when the platform can natively support containers without having to deploy a series of resource-consuming VMs. There are platforms that provide native container support on a bare-metal OS. Datrium, for example, lets you run containers on a bare-metal Linux instance.
Other platforms provide other benefits. Nutanix provides Container Storage, which offers persistent storage for containers running elsewhere in the cluster. Similarly, Cisco's HyperFlex 3.0 software features its FlexStorage driver, which enables persistent storage on that platform. If you run containers, make sure you have an in-depth conversation with your selected vendors about how they support containers.
Some HCI products rely on other methods to provide container support and do not support them directly. For example, Pivot3 and Scale Computing can support containers but only inside VMs running on the platform. These products do not provide native support for container-based workloads.
Support. The support option you choose should match your business needs. Do you require 24/7/365 support with immediate replacement, or can you live with next business day support? You may choose to slightly overengineer your environment to be able to pay for a lower tier of support. Doing so will protect you from the failure of a single node.
Finally, the support program you choose may actually need to be multifaceted. If you buy a software-only HCI product, you may need to buy separate support contracts for the hardware and software. Make sure you do not overlook that step.