Hyper-converged infrastructure market isn't right for every shop
Last week, Maggie Jones published an article on SearchConvergedInfrastructure.com titled “Hyper-converged infrastructure market isn’t right for every shop” (http://searchconvergedinfrastructure.techtarget.com/ehandbook/Get-hyped-for-top-hyper-converged-technology-use-cases). I agree with the assertion in the title and the points she raises, with the clarification that they are based on assumptions that apply to vendors who only sell an integrated HCI appliance – a predefined, proprietary bundle of hardware and software - and not to software-defined HCI products. While Nutanix, Simplivity and others tout their products as “software-defined”, the reality is you can’t just purchase their software and load it on your own hardware; you have to consume it as an integrated server/software bundle. The article concisely exposes the main drawbacks of this bundled approach. Bundling HW/SW decreases flexibility, obscures costs, and reduces customer choice. When you decouple the hardware from the software, the customer wins. This is a key pillar of NodeWeaver’s software-only approach.
The first point the article raises is “Businesses considering a hyper-converged infrastructure systems deployment must determine what will happen to the older hardware that HCI will replace […] to establish whether ripping out the old to make way for the new is viable from a cost perspective”. Since NodeWeaver is “Software-defined HCI” – HCI software that you can load on your own servers - you don’t have to rip out the old servers, you can just repurpose them.
The article goes on to point out that HCI isn’t a good fit for applications that don’t have the right balance of CPU, RAM and storage resources. For instance, running an application that needs a lot of storage but little CPU or RAM “Wastes the remaining resources”. This point is also a direct indictment of the predefined hardware configurations that most HCI vendors force customers into. Nodeweaver customers provide their own hardware, and rather than fixed-size appliances, they can configure their servers with the balance of CPU, RAM and storage that’s best for their workload. And if their needs change in the future, they can easily add more. Furthermore, if you want to get creative you can configure a Nodeweaver cluster with “Storage-centric” nodes that contribute only storage, and “Compute-centric” nodes that run the virtual machines.
The final point the article makes is that smaller companies might find it hard to cost-justify the large starting capacities that an HCI cluster is capable of supporting, considering that the entry point for the leading HCI vendors is capable of easily running 50 or more VMs. Customers with smaller demands, whether the use case is a small company, a remote office, an edge use case, or dev/test, often find it difficult to cost-justify over $100,000 for a handful of VMs (while some vendors offer lower cost versions of their technology, those are feature-limited and non-upgradeable). NodeWeaver gives these use cases a cost-justifiable alternative. As an example of what is possible at that end of the spectrum, you could configure a 6-core Xeon-D system with 64 GB RAM, 500 GB SSD and 6 TB raw disk for under $2,000 each, which means you could have a very capable entry-level 3-node NodeWeaver cluster for about $485/month (amortized over 4 years). For many of these smaller/edge use cases, HCI would be very useful and cost-justifiable at $485/month, but not at over $2,500/month. And circling back, if you wanted to repurpose servers you already own, you can license a 3-node NodeWeaver cluster for as low as $300/month.
If NodeWeaver’s software-defined HCI model can open up new use cases for you, we’d love to talk. Please visit us at www.scalewize.com to learn more.