Assembling your own hyper-converged platform is no small task. That said, the DIY approach can still offer a number...
of important benefits.
Vendors sell appliances that let organizations stand up hyper-converged infrastructure (HCI) quickly and easily, but these appliances come with hefty price tags and do not always offer the flexibility companies need to support their specific workloads. As a result, your IT team might decide to take a do-it-yourself approach and build your own hyper-converged platform. You can use commodity hardware or repurpose components you already have on hand.
Building your own hyper-converged platform has a number of advantages in terms of flexibility, but it also has challenges.
What's wrong with prebuilt HCI?
HCI provides a data storage platform that integrates compute, virtualization and networking resources into a unified infrastructure. The typical hyper-converged platform is made up of multiple self-contained appliances clustered into a single resource pool. Each appliance or node includes its own direct-attached storage and is configured with HCI software that integrates resources. This integration makes it possible to distribute storage across all nodes to provide a unified, high-performing storage platform.
Appliance vendors typically offer only a handful of models, and they have predefined hardware configurations to support specific workflow types. This leaves few options for special configurations. Once you've selected an appliance, you have to live with its limitations. And if you go with a prebuilt hyper-converged platform, you're locked into that system's configuration, which can lead to over-provisioning your environment if you end up purchasing resources you don't need.
Appliances also lock you in to the vendor's linear scaling model. When you need more power, you must purchase whole additional nodes. What if you need more compute capabilities, but not storage? Again, you could end up over-provisioning your environment because you must purchase another whole appliance.
The wonderful world of DIY
The advantages of building a hyper-converged platform yourself can be summed up in one word: flexibility. You get to decide how to combine and configure resources based on your specific requirements. Although you must still ensure hardware and software compatibility, you get to select the server, hypervisor, network, storage and other components. This lets you put together a system that best suits your business needs. You can scale components independently of each other. You can add memory, swap out disks, upgrade CPUs and still scale out by adding complete nodes.
This granular scalability is thanks to the ability to use commodity hardware and not be locked into the types of custom-engineered systems many vendors sell. With commodity hardware, you can build a hyper-converged platform with less investment, replace and scale up components when necessary and reuse the parts for other purposes when it comes time to decommission them.
You can also reuse components from other systems when putting your hyper-converged platform together. For example, you might have recently invested in high-performing storage for a platform you can no longer use. Rather than losing that investment, you can repurpose the storage in your HCI stack. The more components you have on hand, the lower your initial investment will be. And with enough components, you might only need to purchase the HCI software. Several vendors sell HCI software as a stand-alone product, such as Maxta's MxSP software.
The DIY approach also has the advantage of avoiding the type of vendor lock-in you get with an appliance. Although no system can avoid lock-in altogether, the level you get with a DIY option is significantly less than what you get with an appliance. In addition, the DIY approach makes it easier and quicker to implement new technologies. When you purchase an appliance, you're stuck with that hardware, and you have to wait on the vendor for software and configuration updates. That's not the case with a DIY hyper-converged platform. You can update and replace components at any time. Plus, if you're using commodity hardware, it's that much easier to slip in new technologies when they become available.
The downside of DIY
The DIY approach to HCI can certainly save money by avoiding the over-provisioning and lack of granular scaling you get with appliances. In addition, being able to use commodity and repurposed hardware can also add to the savings. But estimating the total cost of ownership for any new system can be tricky, especially with an HCI stack you build yourself. In fact, building a hyper-converged platform can actually cost much more than an appliance-based system when you consider hardware costs, software licensing and the human resources required to implement and maintain such a system.
When you build HCI yourself, you must ensure all the components work together correctly. Achieving this level of compatibility can require far more staff hours than you might anticipate. You must research components, carefully assemble them, optimize their settings and perform extensive testing that covers every workflow scenario. You must then put the system into production, a process that can reveal must-fix snags that may not have been evident during the initial testing.
After everything is up and running, you must then perform the maintenance necessary to keep it that way: update and patch software, replace components when necessary, implement and maintain disaster recovery, and perform a wide range of other tasks, testing every step of the way. To make this process even more challenging, you must address these issues without the type of centralized vendor support you get when you purchase an appliance.
A large organization that has sufficient resources to implement and support a homegrown hyper-converged platform might do well with the DIY approach. A smaller organization that cannot meet these demands might be better off with a prebuilt appliance. Before you decide to build your own HCI stack, make sure you fully understand what you're getting into; this is no place for guesswork.
Do-it-yourself DR vs. using the cloud
How to DIY records management
Should you DIY your cloud?