3 Stumbling Blocks to Hyperconverged Infrastructure Success


Gone are the days where IT managers show interest in the traditional stack for most of their workloads. After all, when you walk into a data center and see all the blinking lights, you’re reminded that those lights represent lots of issues that you need to attend to. From connecting the numerous management interfaces; to scheduling training for those separate products; to maintaining all the connective tissue between them; then making them all work together; and knowing who to call when they don’t work together — like for instance when security patches need to be updated, when users experience performance issues or even when a simple bolt comes loose.

Oh, don’t forget there is still this little thing called “the business.” The reason we have all of this gear in the first place. Your organization wants to make money. If the business isn’t making money off of all of those blinking lights, you had better update your resume.

HCI to the Rescue

This is why I never meet with a customer today without the topic of hyperconverged infrastructure (HCI) coming up. HCI is simple. Think of an appliance that runs your virtualized workloads including compute, storage, management and virtualization — all in a scalable, software-defined 2U appliance. You can add additional appliances as your business needs change. Forget the days of sizing a storage array for three to five years of growth. That was a joke and IT almost never hit the mark. That’s because IT typically has no visibility into long-term business needs or changes. We are almost always the last to know. So how can we realistically expect a customer to size an environment for years of growth blindly? This is why HCI hits the mark. You only buy for what you need today.

Unlike years past, customers today are very informed. They have access to YouTube tutorials, tech magazines and everyday software experts like me, who provide useful insight to help reduce all the complexity that has plagued the data center since the client-server world came into existence.

Yes, I said it. The client-server world of today is complex and disparate. The products are all over the place. The mainframe world is far less intricate. Its workload existed for the purpose of running the business. The mainframe has not gone anywhere regardless of what some vendors have tried to tell us. And we can learn something from an environment that is focused on business workload and not all the trappings of a complex data center. This is what hypeconverged promises to deliver again.

Caveat Emptor: Buyer Beware

Regardless of the HCI manufacturer, reducing “stuff” should be its sole purpose — to discard the connective tissue between components or storage arrays, and to migrate away from complex and power-hungry blade server enclosures.

Just like you, I meet with lots of different manufacturers. I hear a lot of jargon, lip service and buzzwords like flash, software-defined, scale-out and deduplication. But the appliance promise of HCI should be to simplify the data center experience. If their product doesn’t do that, I stop listening.

HCI diminishes the number of blinking lights in the data center. That is a tangible way that complexity is reduced. If a vendor cannot articulate that then I recommend they do something else for a living like become a professional Donkey Kong player. That’s in vogue these days, and more in line with thinking from a bygone era.

For all of its benefits, some customers still don’t get all of their money’s worth out of HCI. If that’s the case for you, here are three reasons HCI isn’t performing as expected for your organization:

1. HCI is not for every workload.

Imagine that! But workloads that don’t fit this model are the exception, not the rule. Where we have seen HCI to be the most transformative in an organization has not been in a full-scale, swap-out of traditional gear for hyperconverged gear in the data center. In fact, I recommend you don’t do that. Full scale data center swap-outs are just not something we see. In fact, that goes against the “buy it as you need it” benefits of HCI.

We see success when a customer picks a project or particular workload that is tailor-made for hypeconverged. That’s where the initial investment starts and the benefits begin to show themselves. Then, when the next tray of disks needs to be purchased or the next server needs to be built out, you can just expand the hyperconverged cluster.

If you are experiencing problems with HCI ask yourself “Is it really a performance problem?” No. Well, maybe it’s a licensing issue. For example, when we talk to customers about HCI and Oracle comes up, we focus on exactly what the workload is and build a small custom environment for that. This usually includes a small all-flash storage array that can take advantage of deduplication and smart snapshot capabilities. Then we move on. There is no reason a single workload should ever hold an organization back from real change in the data center.

2. The vendor sold you on enhanced hardware — and not software.

This is a world of commodity hardware and software-defined “everything.” The minute a company introduces a card, a switch, an ASIC or an “anything” that takes a customer away from commodity off-the-shelf-HW and a SW-defined stack, say goodbye.

Ask yourself, “By buying this will there be less gear on my floor?” If the answer is “No,” forget it. I have heard it all: “You need this switch because it allows us to scale as your environment grows.” Laughable, man. “You need this hardware acceleration card because it makes us fast and we can dedupe.” No you don’t. Everything is fast. This is 2016.

The IT graveyard is littered with headstones of great gear that died on the vine. Why? Because they created the same kinds of maintenance headaches as before you switched to HCI. HCI has been successful in customer environments when it has allowed them to get over the wall of infrastructure. That’s it. Whether it’s at the core of the data center or at some remote office, reducing cabling, devices, connective tissue and multiple management interfaces has been proven to be the key to increasing actual scale and performance. All of that gear supports untold levels of complexity and cost.

Rebuke any new hardware vendor hitting the scene with claims of super-fast and super-scale. Everything is fast. Everyone is riding the Intel “CPU curve.” Everyone buys the same drives. The secret is in the software stack. Claims of scale or performance based in hardware are a thing of the past. End of story.

3. You relied on the vendor’s advice alone.

Success happens when you have a trusted advisor that does not have an agenda. What will reduce complexity in your IT setup? Only you can answer that? Whether someone is promoting a product or not, they should be helping you ask important questions like, “Does this product have a proven track record?” “How many blinking lights can we remove from the data center?” “What will I no longer need to manage and support?” —or more important — “By buying this, will there be less gear on my floor?” If the answer is no, forget it. If they cannot articulate how this product will make life simpler in the data center experience, I strike the gong very quickly. Exit stage left!

So listen up storage and compute companies: regardless of claims for scale, performance or ease-of-use, if you are not reducing the gear in the racks, and reducing routine maintenance for IT staff, I won’t talk about your products to the folks that trust me. I’m not riding your bandwagon, and your agenda does not interest me.

For more information about Hyperconverged Infrastructure watch this video or set up a meeting with your CDW Aggregation Integration and Managed Services Specialist.