How to buy hyperconverged infrastructure: What to ask before investing in HCI

The traditional data center is built on a three-tier infrastructure with discreet blocks of compute, storage and network resources allocated to support specific applications. In a hyperconverged infrastructure (HCI), the three tiers are combined into a single building block called a node. Multiple nodes can be clustered together to form a pool of resources that can be managed through a software layer.

Instead of a server with 50 cores, 128GB RAM and 1TB of storage, you can have 500 cores with 1.2TB RAM and 10TB of storage across 10 nodes, presented as a pool of resources to mix and match into services that deliver the specific performance characteristics and back-end resources needed for the job at hand. Configuration can be done on the fly, through an easy-to-access interface that lets you build or scale your solution.

The result is that you get better utilization, which eliminates the need for overprovisioning and enables a better total cost of ownership (TCO). Compared to the traditional three-tier architecture, HCI also is typically more compact and consumes less power.

Part of the appeal of HCI is that it combines storage, computing, and networking into a single system to reduce complexity and streamline deployments across data centers, remote branches, and edge locations. But be aware: A bundled, modular approach can provide simplicity at the expense of configurability. It’s important to consider issues such as compatibility with legacy infrastructure, scalability limitations and support for cloud tie-ins when you’re choosing a platform. This guide will help buyers of hyperconvergence technology navigate the decision process and provide key questions to ask potential vendors.

HCI appliance or software?

Your approach to hyperconvergence will be tied to your existing investment as well as your future plans. Are you buying all new equipment, or do you need to leverage existing infrastructure? Do you have special workload requirements to address?

For enterprises looking to invest in HCI, there are many questions that need to be answered either internally or in discussion with vendors or outside consultants. Among the first choices to make is form factor: Do you want an HCI appliance or a hardware-agnostic software-based solution?

A hyperconverged appliance will provide preconfigured nodes of compute, storage and network resources, packaged in their own chassis. Some leading vendors that sell HCI as an appliance include Cisco (Hyperflex), Dell EMC (VxRail), HPE (SimpliVity), Scale Computing (HC3), Pivot3 (Acuity), and NetApp (NetApp HCI).

Each HCI node includes compute, storage, and networking resources—the bundled nodes are the building blocks of the infrastructure. Buyers can make certain configuration selections: You can purchase nodes configured with all high-performance SSDs, for example, or choose GPUs to address specific workload requirements. Scaling capacity is as simple as adding additional nodes to the appliance. Once part of the appliance, you can logically assemble systems for your specific performance characteristics through the orchestration software layer.

One advantage of appliances is they let you to choose fully integrated and performance-tuned hardware right out of the box. Plug them in, turn them on and deploy as you need. For growth, you add more building blocks to the appliance to grow the pool. The appliance vendor assumes responsibility for making sure all software and firmware updates and patches have been tested and certified in advance of deployment, easing the burden of your staff having to track and test across multiple vendors. Downsides of this approach can include vendor lock-in and difficulty integrating existing data-center components.

If you already have a three-tier infrastructure that you want to continue to use, you may want to consider a software-based approach, available from vendors including Nutanix (AOS) and VMware (vSAN). This requires licensing a suite of applications that provide a hypervisor, storage-management, network-management and orchestration software to provision and manage your server, storage, and network components. This approach is hardware agnostic; you can use any hardware so long as it meets the firmware revision requirements of the HCI software suite and is not so old that the software no longer supports it.

A software approach allows you choose and maintain the hardware you want from the vendors you want, creating a truly heterogenous environment. A downside is having multiple vendors to deal with for support, which places greater responsibility on your IT  team to resolve integration problems, track bugs and coordinate patches.

Scalability of HCI

The building blocks of HCI are nodes of compute, storage, and network capacity that use virtualization for configuration and an orchestration layer to allow administrators to manage them as a single pool of resources. It’s important to understand the minimum requirements and maximum size of these building blocks in order to gauge the cost of future growth.

For example, HCI typically requires that to grow your storage pool, you need to add nodes with compute and network capabilities. Some nodes are limited to a specific type of storage—HDD vs. SSD, for example. If you wanted to add network-attached storage (NAS) or a storage area network (SAN), you wouldn’t be able to manage them in the orchestration layer and would have to connect them individually to the virtual servers you create.

When thinking about scaling storage, you may not want to add additional CPUs and network ports. Some vendors of either appliance or software-only solutions support disaggregated HCI, which allows the addition of external storage devices such as a SAN into the mix so that you can grow your storage capacity separately from your compute. This is important if you anticipate the two growing in dramatically different proportions or have special performance characteristics you need to address (databases, for example).

Another scalability issue is overhead. The layers of software that abstract the hardware and allow HCI to work need to be considered as you plan for capacity and expansion. Ask your vendor how much overhead to account for.

In addition, make sure you’re aware of the upper capacity bounds of the orchestration software. Will you hit a point where you can no longer expand and be forced to create a new pool of infrastructure? Under those circumstances, you will be forced to manage the infrastructures separately and lose some of the benefits of hyperconvergence, such as data protection and live server migrations. Find out if you’ll be able to move services between the two pools or if you’ll need to treat them like separate data centers. 

Finally, what are the limitations of your license pool? How much can you grow your HCI before you need to go back and write the next check? Anticipating additional license costs and fully understanding what they are is critical to understanding TCO. 

HCI orchestration and interoperability 

An important consideration of HCI is whether the vendor’s orchestration layer can support the complexity that comes along with disparately configured hardware. 

Ask your vendors: Will the orchestration software be able to mix building blocks of disparate sizes (for example, servers with different core counts and RAM capacity, different disk sizes, and types of storage technologies)? Or do all the building blocks have to be of the same configuration to be added to the HCI? If you want to incorporate any legacy resources, will you need to upgrade (or downgrade) them before they can become part of your HCI pool? It will add to the bottom line if you need to buy additional RAM or storage for your current equipment.

Understanding how the orchestration layer sees equipment can make a big difference when migrating to an HCI. Will the HCI be able to poll the network and easily incorporate existing equipment into the pool of resources? It can be very labor intensive if you have to move existing applications before you can add the applications’ infrastructure into the HCI pool. Be prepared for extensive and potentially disruptive work if you need to go through a process of moving existing services, then re-initializing and restructuring their underlying hardware into building blocks before the HCI can use them. 

Vendor lock-in can also be a concern. Does the solution allow you to add hardware components from any vendor? By abstracting the physical hardware and placing it into an HCI, you need to know if you can add any equipment, both legacy and new. This allows you to expand as you need to and avoid being locked into any specific vendor. Even if you choose to go the appliance route, you may want to incorporate legacy systems that are acquired outside the appliance. You should know if you can continue to use equipment purchased from other vendors or if you are locked into a vendor’s specific, proprietary solution. 

What hypervisor platform will your HCI solution use? Proprietary solutions make it likely you will need to buy future building blocks from the same vendor. That could be desirable, as it gives you one source to contact when you need support, but it also puts you at the mercy of that vendor for availability of equipment and software updates. Many vendors use a common platform, such as VMware or Nutanix AHV, but typically you can’t mix and match in the same HCI, so an appliance with a customized hypervisor can be an issue. 

Do you have special peripheral requirements? Are you dependent on GPUs, special sensors, or other types of hardware? Make sure you understand if you can build customized nodes and if you’ll be able to manage any customizations within the HCI or if you’ll need to build something outside the infrastructure to support them. 

Hyperconvergence features to ask about 

As you’re evaluating HCI platforms and weighing investment priorities, here are some areas to dig into.

Does the HCI speak natively to public cloud services?

If you need to quickly expand services for a short business event (say you run an e-commerce site and need additional web servers for a special event), can you burst capacity into the cloud or will you need to overprovision within the HCI to support the business peaks? Will your orchestration layer be able to treat public cloud resources as if they are just another node in the HCI? Having seamless access to the cloud will allow you to quickly grow and shrink capacity without the costly and time consuming (and potentially disruptive) process of shutting down services and manually migrating them. Some public cloud providers have worked to streamline integration between HCI platforms and the cloud. 

What are the built-in redundancies and disaster recovery features?

Will you need to provide additional building blocks and duplicate hardware for redundancy? Can you use cloud resources for this application? Will your data be automatically striped across multiple nodes, or do you need to add additional software to make that happen? Integrating an HCI with a cloud service can provide a place for data to be stored safely off premises in the event of a disaster. Being able to replicate your services in the cloud, and then activate them in a disaster will dramatically reduce your recovery time objectives (RTO).

Can backup appliances or SaaS work within your HCI?

You are probably already using some system to back up your data center. Will you be able to attach a tape robot to a node and have it visible through the orchestration layer so you can run a backup system to write to tape if you require? 

What backup/recovery support is built into your HCI environment if you want to use a SaaS-based product such as Druva or Veeam? Can you attach a backup appliance from vendors such as Commvault, Rubrik or Cohesity? This is becoming a more common feature, where the cloud backup is seen as another tier of storage in the HCI. The backup system runs as a virtual appliance on your HCI and speaks directly to the cloud. 

What additional storage features are built in to the HCI? 

Some HCI platforms provide tools that handle data deduplication to conserve storage space or encryption (in transit and at rest) as your data travels through the network for security and compliance. It may be that separate tools are required to do work such as periodic snapshots of your services for quick recovery in the event of a problem. Make sure it’s clear which features are standard and which require additional tools. 

How do you handle specific high-performance requirements?

Bare metal is faster than HCI as it removes all the abstraction layers that add potential overhead to processor and storage access. But bare metal can’t be supported in an HCI, since you can’t control the hardware if it’s not running the abstraction layer software. That’s why it has been common to see large database servers with heavy performance requirements, for example, stay outside an HCI. Disaggregation is starting to solve some of these issues by allowing separate storage, such as a SAN, into your HCI, as well as enabling enterprises to incorporate special hardware building blocks, such as GPUs, to boost performance for specific applications.

How will product support work? 

Do you call a single vendor for hardware support and a different vendor for hypervisor or orchestration support when you buy an appliance? Having support that can look at your problem holistically will bring faster resolution, and nothing is more frustrating than having multiple vendors pointing at each other to solve a problem. 

Zero-day vulnerabilities are becoming commonplace, and they put the entire enterprise at risk. You should find out how your HCI vendors track for these problems and how quickly they provide tested updates to their customers. 

Can you update the components of your HCI on separate schedules or do you have to update everything at the same time? It can be advantageous to patch a portion of the infrastructure and test it for issues prior to performing a complete update. Pushing massive updates across the entire infrastructure is sometimes necessary and often disruptive. 

There are many benefits to building an HCI environment. You will be able to quickly fulfill your server requirements using familiar tools, reduce the time it takes to bring applications online, and ease maintenance requirements. But enabling flexibility upfront shouldn’t add complexity on the back end. Make sure the hyperconverged infrastructure you invest in will serve your application needs today and as you grow in the future.