Top Ten Tips for Storage Hyperconvergence
All Hyperconverged Infrastructure (HCI) vendors claim simplicity of deployment and ease of use. But one man’s simplicity is another man’s tangle of wires. Just ask anyone transitioning from an old flip phone to a smartphone how difficult it can be to do basic things on a touchscreen phone like enter contacts, make calls and arrange settings. After hours of frustration, the usual solution is to hand the device over to someone younger with a desperate plea for help.
To avoid the same thing happening to those introducing HCI, here are some top tips from the experts.
1. Simple is Relative
Kelly Murphy, founder and CTO, Gridstore, said that all HCI is simple relative to traditional infrastructure. But under the covers there are major differences in architectural approach that impact performance and economics.
“Look deeper than the vendor claims of simplicity,” said Murphy. “Take the time to get to know the company and product set under consideration.”
2. Think Appliance
HCI can be implemented on a build-your-own basis, using rack systems, commodity hardware or appliances. There is no one-size-fits-all, but those seeking to minimize set up issues may be wise to look for a plug-and-play appliance.
“Look for solutions which combine full-featured storage and compute into a single appliance,” said Travis Vigil, executive director, product management, Dell Storage.
3. Balance Simplicity with Cost
The convenience of valet parking comes at a premium. It’s the same thing with HCI appliances. There may be a price to pay for having a vendor package everything in one box and ensure it is integrated. Therefore cost has to enter into the equation.
“While all-in-one appliances allow for a more plug-and-play approach and definitely let users scale quickly, sometimes the savings can be undone by the integrated cost of OEM hardware,” said Ken Dai, general manager, Promise Technology.
4. Commodity Market
One good way to bring costs down is to opt for HCI tools that make use of commodity hardware. There are a number of them on the market.
“We suggest tightly-coupled solutions based on commodity hardware,” said Dai. “Choosing simplicity and single pane-of-glass over excessively complex tools that profess to do everything is something to look out for, and will help with reining in unnecessary costs.”
5. Look Beyond Storage
Storage managers evaluating HCI may have too limited a view of its value if they focus rigidly only on their immediate storage concerns. Yes, these are vitally important, but the whole point of HCI is to address compute, networking and storage on one unified platform. That requires the spanning of attention and a broader view of IT.
“Administrators and managers should look at hyperconvergence as a complete on-demand IT solution and not just a storage replacement,” said Yoram Novick, founder and CEO of Maxta. “The full potential of hyperconvergence is unleashed once organizations focus on optimizing their applications by deploying and managing IT entirely at the VM level.”
6. Virtualization Management
Today’s HCI tools are digging deeper into the application space to encompass virtualization management. As virtualization typically involves various management tools and add-ons, some HCI vendors have taken the step of incorporating these functions within their appliances.
“Hyperconvergence of storage and compute is merely the beginning, and solutions should be looked to for replacing external hypervisor management tools, third party virtualization licensing requirements, and third-party high availability solutions,” said Jason Collier, co-founder and chief evangelist at Scale Computing. “This means that all or nearly all IT workloads can be run on the hyperconverged platform.”
7. Start with Upfront Costs
In the beginning, compare the price tag for the appliance or HCI elements to buying servers and equivalent performance storage systems separately. But at this stage don’t consider soft costs or OPEX.
“Some claim to save 40 to 60 percent compared to traditional infrastructure but base that figure on softer operation cost savings,” said Seth Knox, vice president of product marketing, Atlantis Computing. “While there are operational savings to be had, I would recommend pricing out a competitive new traditional infrastructure against hyperconverged systems.”
8. Use Your Servers and Blades
In some cases, aging storage and server hardware is ready for a refresh. But in others, the metal is still gleaning and that new server smell continues to waft through the data center. In that case, it doesn’t make sense to dump the new gear, replacing it with HCI equipment.
“For those that have recently invested in server infrastructure, particularly blade systems, they should consider using a software defined storage (SDS) platform with their existing servers to get the same benefits at a lower cost,” said Knox. “You can add a couple SSDs or HDDs to your existing servers, get an inexpensive software license, and have a hyperconverged system up and running in a couple days at a fraction of the cost.”
His recommendation is for those with servers less than three years old to use what they have and add HCI software over the top of it.
9. Question the Numbers
Almost every vendor can put up a convincing case as to how its gear is best. They back this up with performance numbers, but these should not be accepted at face value. In some cases, the results can be manipulated to put the product’s best foot forward. Those numbers may not translate into the real world. There is a lack of standard, independently verified performance benchmarking tools for hyperconverged systems, at least for now. Therefore, Knox advised storage managers to push for performance data published by independent third-parties and if not available, check the configuration and workload data provided by vendors to make sure it is similar to how you plan to use the product. In addition, conduct testing in your lab with a cross-section of the workloads you plan to run on the hyperconverged system.
10. Avoid Linear Scaling
Some boxes and configurations force you to add fixed increments of storage and compute simultaneously. But what if your storage needs outgrow your compute requirements? The likelihood is that there are many cases where you need to scale storage faster than compute.
“In remote offices and branch offices or disaster recovery deployments, you can sometimes use hyperconverged appliances in one location and make do with lower cost infrastructure (DAS, JBOD or old NAS) in the other, rather than an appliance only strategy for both locations,” said Knox.
Photo courtesy of Shutterstock.