Storage Hyperconvergence: When Does It Make Sense?
Convergence can be great thing: the U.S. states came together to form an entity greater than the sum of its parts.
But convergence can also lead to burdensome creations that accomplish little.
It can be the same with storage hyprconvergence. There may be times when it makes sense to implement hyperconverged infrastructure (HCI), and other times when all it may achieve is to inflate the budget or keep IT occupied with a new architecture.
So when should you introduce it, and in which cases should you avoid it? And if you are embarking on this journey, where should you start?
Avoid Some Legacy Workloads
Large enterprises, in particular, often have old applications written in all sorts of ancient code. Many of these can be problematic for HCI.
“Distributed legacy workloads that require re-writes before they can be virtualized are generally poor candidates for hyperconvergence,” said Ken Dai, general manager, Promise Technology.
Not Just for Small Organizations
A big part of the sales pitch for HCI has been that it simplifies the demands on IT and makes it less necessary to increase manpower. That led small organizations to dominate the early adopter category. But HCI is far from a small business-only proposition.
“There is a common misperception that HCI is only for small or hyper-consolidated environments,” said Greg Schulz, an analyst at StorageIO Group.
Start with an Inventory
It is best not to rush into HCI or get caught up in the hype. Take your time to inventory your overall IT needs, assess the situation carefully and see if you can find the right fit.
“Some older workloads and job functions may require traditional hardware solutions,” said Seth Knox, vice president of product marketing, Atlantis Computing. “Try implementing hyperconverged approaches gradually to show results in specific parts of your business and job functions.”
One example might be to deploy a hyperconverged all-flash storage solution as the foundation for a Windows 7 to Windows 10 virtual desktop migration. Once that is rolling and showing good results, look at other areas for HCI to expand into.
We’ve been hearing a lot about Software Defined Storage (SDS) and the Software Defined Data Center (SDDC) for a couple of years now, but some organizations are either hesitant to adopt it or are unsure where to start. HCI may be the smart way to introduce it gradually into the enterprise.
“The best ROI cases we have seen from customers involve using SDS first to make their existing SAN- and NAS-based storage infrastructure more efficient,” said Knox. “This eliminates the need to buy additional disk shelves or arrays and then you can migrate over time to hyperconvergence either using an appliance or software approach over a three- to five-year period.”
Most IT shops have heavy investments in existing equipment as well as outstanding maintenance contracts. Therefore, a good time to consider adding HCI is around the time when those contracts expire.
“We’ve seen many customers end of life all their SAN and NAS storage arrays at the end of their existing maintenance contracts,” said Knox.
For those on extended contracts and with a lot of money tied up in the data center, a good way to ease into HCI is to use it to upgrade the infrastructure of branch offices or disaster recovery (DR) sites initially while still using existing compute and storage assets in the primary datacenter, said Jason Collier, co-founder and chief evangelist at Scale Computing.
“It may even make sense to use hyperconvergence initially for individual departments or for dev and test before deploying it across the entire datacenter,” he said.
Compare to the Cloud
There are plenty of comparisons out there about how HCI measures up against traditional storage and data center technology. But that may not be the best way to view things.
“The new point of comparison for delivering IT services is public cloud solutions such as Amazon Web Services (AWS) and Microsoft Azure,” said Prabu Rambadran, director of product marketing, Nutanix.
As well as running the numbers of HCI versus SAN/NAS, it would be wise to compare it to offloading everything to the cloud. Some workloads may work better in the cloud, while others will be good candidates for convergence.
Hyperconverged infrastructure has matured to a point where almost all virtualized applications can now be supported. But most users are hesitant to dive in. They typically prefer to experiment a subset of their workloads first before moving all their applications.
Where to start? Rambadran recommended moving some general-purpose server virtualization workloads, Virtual Desktop Infrastructure (VDI), dev/test, big data applications, Microsoft Exchange, SharePoint and/or remote offices and branch offices (ROBO) to HCI as an early step.
“Evaluate how the solution performs and then move other mission critical applications built on Oracle SAP, Microsoft SQL etc.,” he said.
At least for now, HCI isn’t invading the top end of storage in terms of overall scale. So leave that area well alone until the technology matures even further.
“Most hyperconverged offerings are not suited to large-scale storage requirements such as object stores or storage in the tens of petabytes range,” said Kelly Murphy, founder and CTO, Gridstore.
That said, some hyperconverged systems have found success within mid-sized environments because of their ability to eliminate complexity, expense and the latency sometimes associated with SAN-based storage solutions, added Murphy.
“Hyperpconvergence, by definition, collapses the separate elements of the data center — the networking, compute and storage and in one offering — and provides the main elements that an onsite datacenter requires,” said Murphy. “By virtue of the economical footprint, hyperconverged offerings are more frequently becoming the go-to investment for companies looking for a data center refresh.”
Grade Upwards Gradually
But these points of advice may not hold true for long. The external storage market is seeing a sales decline. Even the relatively new all-flash array market is being disrupted by all-flash hyperconverged offerings. It may not be that long before an all-flash datacenter becomes realistically affordable for the average enterprise. So that could see hyperconvergence invading everywhere.
“We see hyperconvergence expanding to all types of workloads including critical workloads that need high levels of performance and reliability,” said Murphy.
Rapid Scaling and Time to Deploy
Pressing immediate need is another area where HCI might offer up an answer.
“The quickest ROI is achieved when deploying HCI systems in application environments that need to rapidly scale a virtualized environment, and where the priority for the user is time-to-deploy,” said Dai. “HCI systems allow the user to reduce the cost of storage scale out.”
Photo courtesy of Shutterstock.