Is HPE Nimble dHCI HCI 2.0?
Transcript
(upbeat music) >> So HPE gave the CTO Advisor, as a loner, a Intel-based dHCI system to put their marketing claims to the test. Hmm. That's a bold move. Let's see how it turned out. So in this three-part video series, we are going to take a look at HPE's marketing claims, what we need as data center operators to deliver business critical services, and see if HPE lived up to the test. We're going to break this series into three different videos. Video one, you're watching now.
We're going to talk about the architecture and the why of dHCI. Video two, we're going to talk about the management capabilities of the dHCI solution. And then video three, we're going to talk about real workloads and use cases and how we put that to the test in the CTO Advisor hybrid infrastructure. But first let's level set. 0, the how and the what of it. And then I'll dig into our experience with the initial setup and config of dHCI. But more importantly, why should this even matter to the CIO, CTO, et cetera.
Isn't this a solved thing? So let's first talk about the problem, the CTO Advisor specifically, why we asked HPE to give us a dHCI system to compare it to what we're doing today. First off, this is I think a big part of the architecture. Let me grab my little iPad here and draw a picture for you. What we have today is what Tom would call ice cream inside of our cake. We have our slice of cake and inside of that, we've placed ice cream.
And if you like ice cream and cake, and you only need one helping of each, ice cream is out of the cake, brilliant idea, HCI. If you only need one piece of the whole thing, then you're good. HCI vendors, since the onset of HCI have been working to solve the architecture problem of scaling linearly, both storage and networking, and compute. HPE believe they have solved that with dHCI. Video two, we'll dive into whether that's the case. The second operational challenge that we had in the CTO Advisor hybrid infrastructure is that of, again, operations related problem.
When we need to expand our current HCI solution, it's quite a bit of work. 0. dHCI promises to eliminate that need. I think we experienced some of that in the install and the initial deployment of HCI, I'll share that in a sec. All right, Chuck, thanks for joining the video. We're going to hit it right off the bat. 0. First, tell me why is HCI broken and why did you guys think that you have the solution to it?
>> Yeah. Well, Keith, great to be here. Just excited to talk about this. So in our point of view, what customers have been challenged with. Well, first of all, the opportunity with HCI is terrific. The customers love the simple experience they get with a hyper-converged infrastructure. It's really making their lives simpler, to simplify the complexity in IT. The challenge is they want to bring this now, this experience, to more and more workloads. And what they're finding, the architecture of traditional HCI, and I love the analogies you've been using, about it's all in one, it's one building block.
As you scale it, and you start to run more and more workloads, more and more mixed workloads, large VM, small VMs, you start to see some issues with the architecture of scale-out HCI. And it deals with, we call it dealing with unpredictable workloads that you don't know how many large VMs you're going to have a month or not six months or nine months from now, or what are your workloads from database to three-tier apps that you want to put on this new HCI infrastructure or capacity driven apps, like security and video capture, all of this needs to run.
So what we found is to get the resiliency, the availability that you want, you want to set the infrastructure and let it run. You want the performance, literally, sub-millisecond latency performance, and you want flexible scaling to deal with this unpredictability. Those tenants aren't really built into the HCI architecture and you're sort of left with having to make a lot of trade-offs to get that performance, resiliency and scale that you want. And many times the products are falling short. So over a year ago, HPE set out to fix these issues and overcome these trade-offs that customers are having to make and give them a similar experience.
>> So, Chuck. How is this architecture different than HCI that we see in traditional vendor deployments? >> Right, right. And to address this, we have incredible innovation to bring to bear for our customers and our partners. And you hit on a couple big ones. HPE ProLiant, just one of the best engines for running virtualized workloads in the world are non full-stop. HPE ProLiant from all the generations of servers, we're up to gen 10 with tremendous innovations. It's great, it's a great compute memory tier.
And we have great storage architectures like Nimble Storage. Well, what we've done is we've taken those two-piece parts and combine them with a tremendous amount of automation. And you've hit on many of these items, I know in your evaluation. And we call this HPE Nimble Storage dHCI. It's a total solution providing that HCI experience with that automation. So right from VMware Vcenter, you can manage your VMs. The verbs are, I want to add a VM, I want to delete a VM, I want to back up a VM, I want to monitor, maybe grow the VMs, expand them, or even shrink them.
And then I want to deliver performance through them and policy automation for backup. That's what customers want in HCI, manage the VMs and the apps, not the compute necessarily and the storage. And we provide these incredibly powerful piece parts in the architecture. And that's what comprises HPE Nimble Storage dHCI and it gives customers the resiliency they want. Literally, six nines data availability because of the architecture. We can survive three simultaneous drive failures and keep your apps running. As well we have the performance.
Sub-millisecond latency for performance to run these workloads and then do it in a way that you're not colliding VMs through HPE info site. In our analytics engine, we show you visibility to make sure your apps keep running, even as you're running more and more apps and you're getting the latencies that you need for you to run your business, not just VDI or tier-two apps, but your business critical apps. And then finally the flexibility, the new thing, that we really revolutionized is you can now scale, add more ProLiant compute nodes, literally, in a few minutes automatically.
It's all done for you, non-disruptively to your VMs. Or you can add a storage through the Nimble array and that gives you a flexible way to address your needs as a business, much more efficiently than traditional HCI, where you may have to add a compute element and a storage element. And in some way you may be over-provisioned. If you just need compute for apps, why bring in more storage that's not necessary? Or similarly, if you just need storage for your video or your databases that are extremely large or TestDev, why bring in more compute that could add extra tax on the environment?
So that's what's novel about the architecture. >> So you just heard from HPE their perspective and why they did dHCI, what is dHCI and why they believe it fits in the marketplace. Let's talk about the CTO Advisor and the business value of dHCI for us in our environment. IT is cumulative. You've heard me say that before. If you've deployed HCI in the past, you know that HCI more or less has not replaced your infrastructure. You still at the very minimum have public cloud, your HCI and probably your traditional storage compute and networking stack.
Your staff can only do so much. So while HCI is yes, simple, it is, again, another platform to manage inside of your data center. The promise of dHCI is that I'm taking components that operationally, I already know how to manage. I know how to manage a Nimble Storage array from a hardware perspective. I know how to manage DL360s and DL380 ProLiant servers. I know my Intel platforms like the back of my hand, why take a separate platform and put it into my data center while at the same time I need to learn public cloud?
On top of that, I'm asked to facilitate machine learning and AI. I need to get business decisions faster, and I need to support all of my legacy infrastructure and applications. Does HPE dHCI allow me to gain this business capability will fall overburdening my staff? Let's dive in and take a look. Let's first look at legacy storage compute networking, that install process. If you built infrastructure over the past 20 years, it hasn't changed much. We take our storage arrays. (popping sound) Build a storage network, take our compute.
Build the parts necessary for our compute, memory, CPU, local disc. And then we take our networking and we run that independently. These three silos is something that we're accustomed to managing in an enterprise data center. There's nothing 15-minute about that. When it comes to managing and deploying infrastructure, it takes days, weeks, months. The first claim that we'll look at is that you can deploy dHCI in 15 minutes. Wow, that's quite the claim. First off, it takes more than 15 minutes to rack and stack dHCI.
So one of our first kind of back and forth with HPE was where do we start the clock? And where do we end the clock? The fundamentals were that we had our infrastructure laid out. Storage was, I'm sorry. Networking was already in place. The assumption is the physical servers are rack and stack. And now I'm planning to install VMware vSphere and provision the storage. Okay, fair enough. We get it to that point, hit go, and honestly, we ran into problems.
What problems did we run into? Well, first off, HPE in their efforts designed decisions that weren't in line with the design decisions that we made. So we went back and rerun the test based off of HPE's design assumptions. And it deployed in 15 minutes, give or take. We ran tests a couple of times. Give or take a couple of minutes here or there. The 15 minute claim is accurate, but what insight do we get out of the 15 minute claim?
One, again, you can't bolt down weeks and months of effort into 15 minutes. However, you do have to do more with less or the same resources that you have at the time. And I think the aim of the claim is to give insight into that capability. How long did it take us practically to do this? It was less than a day. It took Tom by himself, with some troubleshooting, less than a day to deploy dHCI custom to our environment versus the out-of-box experience, which assumed greenfield.
So, you know, HPE, you get a check on some ease of simplicity for deployment, but 15 minutes, you know, not many environments are 15 minutes. So what are our initial thoughts around the architecture? Traditional architecture is what Tom likes to call having your ice cream inside of your cake. If you want or desire a single serving of ice cream and cake, that's great. If you want to change the flavor of your ice cream, but keep the flavor of your cake, you have a problem.
If you want to change the portions of it, you have a problem. Practically, in the data center, that's what happened with us with a vanilla HCI deployment. It was rigid. We had trouble going from one version of the control plane to another version of the control plane. dHCI, the architecture, allows us flexibility to choose the flavor of storage, the flavor of compute. It allows us basically that traditional sense. Is it the day two experience that we expect and will it lead to the types of capabilities that enterprises need to deploy today to get the technical agility that they need to support the business objectives?
Well, we'll find that out in videos two, which will tackle management, and video three, in which we talk about what it was like to run workloads on dHCI. Stay tuned. com. Talk to you next video.
