Hyperconverged infrastructure has emerged as one of the hottest data center trends in recent history. With a lot of promises and a lot of hype, it’s important to understand what hyperconverged infrastructure is and what it isn’t.
At its most basic, hyperconverged infrastructure is defined as a conglomeration of storage, compute, and a hypervisor, all neatly packaged on a single server, with no separate SAN or storage array needed. All of the storage resides locally inside each server. At first glance, many people look at hyperconverged infrastructure and their initial reaction is, “Why would I want to go back to the days of direct-attached storage?” However, while it’s true that hyperconvergence does bring you back to those days by fully eliminating the stand-alone storage layer, it does not bring back the downsides inherent in that approach to storage.
Unlike the old days in which all of that storage was trapped on individual servers, hyperconverged infrastructure binds it all together with a powerful software layer connecting all of the nodes over a standard Ethernet network.
In essense, you’re binding all your storage and creating a distributed, highly available storage fabric that’s fully integrated with the server and hypervisor components. Hyperconvergence brings together two key trends:
- Software-defined storage
- Use of commodity hardware
People sometimes confuse software-defined storage (SDS) and hyperconverged infrastructure because there is a lot of similarity between these two approaches to storage. In the world of software defined storage, it’s sufficient to have a system that abstracts storage hardware resources and aggregates those resources for eventual consumption by connected hosts. The aggregation process is handled by a software-based controller of some kind.
Hyperconverged infrastructure can be considered a subset of software-defined storage in that there is a software-based controller of some kind (we’ll discuss this later in this paper) which effectively abstracts the physical storage layer, and presents it to a higher order construct. In the case of hyperconverged infrastructure, that construct can be a hypervisor (for in-kernel storage management) or a controller virtual machine that is assigned the task of managing local storage on an individual node. The key difference between SDS and hyperconvergence is that this next-level construct also operates on the same hardware as the storage. This integrated storage component is one of the major selling points of hyperconvergence.
In non-hyperconverged infrastructure software-defined storage systems, the actual storage devices look a lot like traditional storage arrays. You create connections from host machines using your protocol of choice, and the storage layer places data on correct nodes behind the scenes. Meaning, there is physical separation between hosts and storage.
The Role of Hardware
You may have noticed that we haven’t said much about hardware in this new world order. That’s intentional. With hyperconverged infrastructure, hardware is actually a secondary consideration. You still need to have hardware that can perform well, but you don’t necessarily have to go out and buy from a specific brand. Instead, for the server-based nodes in your hyperconverged infrastructure cluster, most hyperconverged infrastructure vendors provide multiple server vendor options from which to choose.
Many equate this deemphasizing of hardware as the hardware being unimportant. That is simply not true. Hardware choice is critical no matter what data architectural option you select. The difference here is that you can take a commodity, off-the-shelf approach to hardware and you don’t generally need to buy very specific brands. However, the hardware you choose still has to be able to perform and has to be stable. You can’t just take that old IBM PCjr you’ve got stowed away in the closet and turn it into a node in your new hyperconverged infrastructure cluster.
By conglomerating servers and storage and throwing in a mix of software, hyperconverged infrastructure has become a powerful way to add simplicity to your data center while enabling the ability for your company to easily scale as its needs grow.