Hyperconverged infrastructure: Speed, flexibility, performance in a box
Hyperconverged infrastructure is getting a lot of attention these days, offering businesses a simple, reliable, and highly flexible approach to architecture. Essentially a single box you just plug in, HCI departs from the traditional three-tier architecture―server, storage, and networking―enabling organizations like Formula One racing to fire up and take down high-performance IT infrastructure in literally minutes.
Chris Middleton, head of IT infrastructure operations at Aston Martin Red Bull Racing, says his team stands up and dismantles an entire data center 35 times a year, and that needs to happen quickly―meaning simplicity is essential. What's more, he says, "Formula One is a performance sport, so … every iteration of our infrastructure needs to be faster than the last. We need to be getting more storage, quicker CPUs, more memory, so we can throw more resources at any given problem."
For smaller businesses with resource-strapped IT teams, an all-in-one solution that's simple to set up, manage, and scale means quicker rollouts of new applications and products―and the ability to better compete, says Matt Harris, vice president of hybrid IT at Hewlett Packard Enterprise.
In this episode of Technology Untangled, Harris joins Middleton, Luke Peters, an HCI specialist at Hewlett Packard Enterprise, and Bageshri Kundu, global strategic product manager for HPE SimpliVity, for an in-depth look at how hyperconverged infrastructure works, its many benefits, and why a hybrid HCI-cloud approach may be the trend to watch in the future.
Excerpts from the podcast, hosted by Michael Bird, follow:
Bageshri Kundu: Hyperconverged infrastructure, or HCI, as we call it commonly, seamlessly combines different components like compute storage, networking, as well as a virtualization layer with the different hypervisors and the data services all in a single physical system.
So the HCI software stack currently runs on industry-standard x86 servers, and those are running virtualized. Or even nowadays, you can see them running kind of what your container workloads do as well.
Matt Harris: You have a single management platform, which therefore takes out complexity. It means that from a hyperconverged perspective, you can have a single administrator who is responsible for the entire stack. You also have an ability to scale in what we call a node format—so very skilled and small building blocks to grow.
Simple, all-in-one solution
Michael BIrd: Converged delivered the three tiers of architecture as an integrated stack, and hyperconverged wrapped all of those with a software layer, simplified, plumbed in, and without the need for mass configuration. This is what IT departments worldwide had been waiting for.
Luke Peters: The key benefit of HCI from an IT perspective … is manageability of the environment. Streamlining the management, having a single generalist, is going to save the organization a lot of time that can be spent elsewhere. Most hyperconverged platforms run upon a VMware environment—so vSphere, for example. If you know VMware, you're already 90 percent familiar with the platform that you're about to invest in. So it's a very, very quick learning curve.
BIrd: So another big plus around HCI is scalability.
Hyperconverged is what's known as a block-based architecture, and so it can be scaled piece by piece, node by node, which means organizations can start small with, say, a single box and they don't need to do a massive architecture design for one project. They can basically scale it out one block or one box at a time.
Harris: Customers are looking to have scalability in line with how that business scales—so not having to overprovision or buy too much architecture, which isn't the right requirement for what their needs are.
Peters: If you just need a little bit more resource, you can scale by just procuring additional nodes, [which can] be deployed into your environment nondisruptively, [with] no downtime to the rest of the environment. It just gets deployed, it assimilates, it configures itself, and away you go.
BIrd: One of the key features of HCI is software-defined storage.
Peters: You look at your three-tiered architecture. You would update your servers independently from your storage, from your fabric switches, and any other software that you have in your environment. That's quite time consuming. Lifecycle management for me is the ability [to do] almost a whole one-click upgrade. In fact, it is a one-click upgrade where you can upgrade that hyperconverged node and update all of the components that are associated with it.
BIrd: Thanks to its manageability, its scalability, its lifecycle management, and quick learning curve, HCI is a pretty neat solution, particularly if you've got a distributed workforce.
Harris: We've seen customers deploy hyperconverged infrastructures in remote sites where they can't have a skilled management or labor resources to manage the infrastructure.
Kundu: Regardless of how widespread the physical resources happen to be, hyperconverged systems can handle them as though they were all sitting next to one another.
So it provides a true flexibility, where you can actually scale up or scale down the sources on demand.
Hyperconvergence in the fast lane
BIrd: Hyperconverged offers portable infrastructure on demand, and who better to talk to us and somebody who literally ships a data center around the world to some of the most un-data-center-friendly environments: Formula One race circuits.
Chris Middleton: We're shutting down a complete data center 35 times a year. So, yeah, which is pretty insane, really, when you think that the average data center probably doesn't get shut down once every five years unless it's for some form of electrical testing,
BIrd: That is Chris Middleton.
Middleton: My name is Chris Middleton. I am the head of IT infrastructure operations at Aston Martin Red Bull Racing.
BIrd: This is hyperconverged infrastructure in the fast lane, where time is of the essence, performance is non-negotiable, and the environmental conditions are challenging, to say the least.
Middleton: In Bahrain, you know, it's 40 degrees. With Singapore, are our servers going to be dripping wet with humidity and shutting themselves down?
BIrd: The trackside team have certainly got their work cut out for them.
Middleton: It's pretty hectic. Traditionally, Monday they fly into the locations; they'll nip into the track just to make sure that all of the freight has been delivered.
Tuesday is unpacking. The garages are in a state where they can just roll the freight into them, racks are unpacked, and everything's kind of cabled up.
And then Wednesday and Thursday is setup and pre-checks, system checks, and making sure that everything's come together in the correct order: Servers are up, status is green, no hardware failures, etc.
Thursday is the day that most of the engineers are rocking up to the track. Essentially, engineers are our customers. We need to make sure that we are set up ready for them to start working, and if they can't work, it means that we're delaying the setup and configure of the car.
So Thursday we need to be almost 100 percent there. Friday [is] double practice. We enforce change freeze as on Friday night, so there are no more uncertainties or adventure or anything like that. Practice three [is] on Saturday morning, and then you're into the sharp end of the weekend and the qualifying. And then Sunday—Sunday, obviously, [is] race day.
And then pack down.
Pushed to the limits
BIrd: During practice qualifying and the race, the data center and infrastructure are pushed to the limits in terms of consumption. But what goes up must come down, and for the team it's got to happen fast, especially if there's a race the next weekend.
Middleton: A race traditionally starts at 2:00 or 3:00 and lasts, let's say, two hours. So we're at 5:00 and then the analysis of all that data then gets done by engineers at the track and engineers in the factory. You've got your mobile data center, which is up and alive while everything is being packed down around it. There will be a final call from the engineers saying, "Right, we need to get this shipped and sent."
And then it's a rush to get the equipment shut down and the data center shut down as quickly as possible without causing any corruption to that data. We can now shut down the infrastructure—I say an hour, but I know that they can do it in 30 minutes now.
BIrd: That impressive 30-minute infrastructure breakdown is thanks to the star of the show: hyperconverged infrastructure—in this case, HP SimpliVity. Everything the Aston Martin Red Bull Racing Team needs to do is at the extreme, but their drivers for using HCI are universal: reliability, simplicity, agility, and flexibility.
But for them, it's just cranked up to 11.
Middleton: Trackside infrastructure is so complex; there's many things that we need to consider. I think the main thing that we're looking at is reliability. If we've got any questions on its integrity or its stability, that could be the difference between us actually completing a race or having data for a lap or two laps, or anything like that.
Next, I'd say we're looking at performance. Formula One is a performance sport, and we're always looking to squeeze out the droplets of what we can for performance. So … every iteration of our infrastructure needs to be faster than the last. We need to be getting more storage, quicker CPUs, more memory, so we can throw more resources at any given problem.
And if we can throw more resource, we can run more iterations of a query or a strategy or a car setup. The more scenarios we can test, the more chance we have of hitting that sweet spot for the setup and therefore the greater the performance of the car.
And then lastly, which we kind of overlook, is agility. … F1 is a really quick-changing, fast-paced sport, and we make our budget requirements, or we initially invest in a solution, for around three years. And looking into how the sport changes, [how] the regulations change, etc., three years in technology is a long time.
So we've got to invest in it, put a big investment into an infrastructure and a platform that's going to last three years without really knowing how the rules of the sport are going to change. We need something that we're able to shift workloads around and make fundamental changes.
It might be that we need more VDIs; it might be that we need less application servers and we need to be doing more GPU-based querying—all that kind of stuff. So we need something that we're able to very, very quickly reallocate resources without spending thousands of thousands of pounds—and quickly, to be able to meet these needs of the business.
F1's a very odd sport as well, where growth rates aren't necessarily linear. Somebody, one of the designers or engineers or developers, may come up with a fantastic idea that suddenly requires, you know, triple the amount of resources that they did previously.
But if it has a positive effect on car performance, then whatever it takes to make that car performance, and being able to quickly and easily reallocate that resource makes that possible, which then directly affects car performance.
BIrd: The ability to reallocate resources when using HCI is directly linked to competitive performance. And, as Matt explains, the benefit really applies to any organization.
Harris: In this day and age, time to market of new applications, new products that customers are trying to bring to market, really does differentiate and give organizations benefits versus the competitors. So speed of deployment becomes critical, and hyperconverged infrastructure—because of the simplicity of it, because of the building block nature of it—means the test and development cycles on applications can be faster.
It means that organizations can bring new products or applications to market more quickly without having to have invested significantly in an architecture where maybe some of those applications or products they're bringing to market aren't going to pass the threshold or the test of market acceptance.
So the business benefits relate around significant cost reduction, simplicity of management by a single person, which gives operational benefit in terms of speed and the ability [to bring] products and solutions and applications to the market far more quickly.
BIrd: As tech rapidly evolves, IT departments are pushed to constantly adapt and often to actually do more with less.
Peters: We want you to run these workloads, or we want you to run this type of environment or add more workloads to the environment, but we're not extending the amount of staff you have in the IT team—just do more. So having that ability to reduce the amount of time managing the infrastructure frees them up to actually go and investigate these new applications, [which] is very important to the business. They don't want to have to worry about the infrastructure. It should just run.
Middleton: I no longer need a network specialist, a storage specialist, a server specialist, a virtual guy—my servers and the storage is all just rolled into one. And that tight interaction between the two means that we're just a lot faster, a lot more tightly knit, and we don't require that same level of highly skilled resource.
BIrd: Aston Martin Red Bull Racing is the extreme on the spectrum of HCI use cases. But as Bageshri explains, moving from legacy three-tier infrastructure to HCI can benefit just about any organization.
Better performance, easier to scale
Kundu: So this retail chain runs 24 hours a day, seven days a week. During the day, the retail storefronts are open for customers, and during the night, their warehouses and logistic operations continue to operate. This particular retail chain needed at least four IT experts to maintain the legacy infrastructure. So doing maintenance on these systems was a horrible experience, right?
They decided to try out HCI due to the flexibility and the simplicity of the solution. So they run thousands of concurrent, kind of virtual desktop sessions, on it for the remote workers. They also run ERP and business continuity applications and critical warehouse applications on virtual machines on top of these HCI systems, and the performance has improved multiple fold. It's a significant jump in their performance, even with a significant load on the system.
So backing up the VMS now just takes a few seconds, and storing these backups—they probably do it like once a month—makes it a seamless process with absolutely no downtime. All this infrastructure can be managed with a single IT resource. And as the business continues to expand, they are able to easily scale that infrastructure to support that growing.
BIrd: Because it's so easy to scale, a lot of organizations have jumped on board with HCI, particularly those who run a large amount of virtual machines. And with the unprecedented events of 2020, organizations that were already using hyperconverged actually saw some of the benefits of that choice in real time.
Harris: We have also seen specific use cases such as virtual desktop infrastructure, which has seen massive scaling requirements, particularly since the COVID-19 crisis that we're all experiencing. And as customers wanting to deploy virtual desktop infrastructure—or VDI, as it's known—they're looking for a far more simpler, cut-down version of how they can do that effectively in that data centers, one which allows for immediate scalability when more workers or users require a VDI-based experience.
Resiliency in a new era
Kundu: It is super-easy to actually deploy a new clusters and spin up virtual machines in an extremely short time. HCI also provides a true resiliency and security for these virtual machines. The new virtualized solutions can also be easily accessed and managed virtually to power up a range of remote applications in the market segments.
Mostly, I would say banking applications, electronic trading—telemedicine is a big, big one today. Supporting remote workers in temporary kind of medical facilities. And also e-learning, where our universities and schools have not been open for a significant amount of time and remote learning is a very important part of that.
So the remote support capabilities that hyperconverged VDI solutions provide, they have been, I would say, very essential to customers.
Peters: Now, if you're not looking at HCI as an organization, you're not really doing your job, because that is the de facto standard now for data centers.
Harris: We're starting to see organizations invest in hyperconverged infrastructure for other use cases, and that could be around data center modernization, enterprise-scale applications, [and] database file and print. So I would say no application is really off the table for a consideration of moving to hyperconverged infrastructure, given the innovations coming out today from HPE and other vendors.
BIrd: For high-octane enterprise IT solutions, HCI ticks most of the boxes. But what about smaller organizations? Well, there is beauty and HCI's simplicity, and for small and medium businesses, it can be a perfect all-in-one hassle-free solution.
HCI for all
Kundu: For folks who are not really familiar with IT, if they actually started looking at traditional legacy infrastructure, it can be quite daunting at times.
And then if they just start kind of digging into hyperconverged, they will see that right upfront, that the simplicity, the ease, it's kind of a no-brainer for those mom-and-pop kind of shops or small retail stores, small gas stations—where the IT needs are very small, the customer does not have a lot of IT expertise or a lot of IT skills, but the organization wants to kind of modernize their IT infrastructure.
They want to start with something small, and they may not really have that larger growth that we see with the other end-device kind of applications. So hyperconverged kind of builds that need very well. These businesses can actually go live overnight. They just need to deploy the small cluster of nodes.
They don't have to worry about other appliances like backups and everything can be managed—that whole data still stays within the data center, stays within that retail store. And I think that's a big plus.
BIrd: Small and medium businesses can plug in and get going right away, but the question remains for organizations at scale. How on earth do you implement when you've already paid a lot for the kit that you got in place?
Shifting from legacy infrastructure
Harris: Organizations no longer need to rip and replace. It's clearly an option, where organizations may look at a new, modernized platform for the future. But organizations may want to leverage investments they've made previously. What we're seeing isa number of hyperconverged vendors really innovate in that field.
An example of that is we have a number of organizations who have made compute purchases in the server layer over the last two years, and they are attempting to move to a hyperconverged infrastructure (dHCI), given all the business benefit. But [with] our platform, called desegregated hyperconverged infrastructure, [organizations can] utilize the server investment they've made previously and bring in the other areas and storage network and the management area to convert their preexisting investments.
BIrd: Whoa, whoa, whoa, desegregated hyperconverged infrastructure. I thought the whole point of HCI was to have everything together in one box.
Harris: It's bringing all the benefits of a hyperconverged solution but taking away a number of the restrictions organizations have shared with us of why they haven't moved to those architectures.
So, dHCI has independent scaling of compute and storage layers. It has a single management interface, which is today based on VMware. It has artificial intelligence embedded within it, which is monitoring and maintaining that infrastructure without human intervention.
BIrd: So, despite sounding like an oxymoron, desegregated hyperconverged infrastructure almost seems like a logical next step. By incorporating AI, it will potentially open up a new world of efficiency and speed.
The trend toward hybrid
Harris: For the last three to five years, we've seen organizations really drive a cloud-first strategy, and that has been driven out of necessity where costs were too much, speed was of the essence, and time to market on products and solutions was required.
The trend that we're seeing now is that the cloud-first only strategy isn't fit for purpose and that one size doesn't fit all. There is certainly an understanding with organization analysts and vendors out there that a hybrid approach to infrastructure and workloads will [have a] prevalent role in the future.
BIrd: So clearly this isn't the end of infrastructure's evolution, and we're probably going to see some big changes in the future.
HCI is more about how our system looks from the outside. It's a simplified box giving you the entire stack. We want it to look integrated from the outside but within the box is totally open for discussion.
dHCI is very much a continuation of that story, offering additional flexibility and capabilities that place it at the cutting edge of infrastructure.
Harris: dHCI and hyperconverged infrastructures classically address some of the enterprise data center needs. But the future of on-premises data centers will also span to what we call the edge. And when you think about technology such as IoT and 5G, organizations will start to reimagine what on-premises infrastructure means.
We'll see organizations look at oil rigs and deep sea and the rain forest in a way that has never been done before or have been possible before. And remote regions that can't classically be accessed or maintained will require technologies which have AI at its core, where they self-manage and self-heal and can collect that data and insight to give business advantage or to solve world problems in the future.
So, who knows from a tech innovation perspective? We know AI will be at the heart of it, and I think you'll also see commercial innovation ramp and continue to be a major theme or a more important theme moving forward.
BIrd: HCI and dHCI are working towards the challenges IT departments are facing now and in the future that involve massive amounts of data and the need for agility, stability, and speed.
You have been listening to Technology Untangled, and a huge thanks to today's guests for joining us: Luke Peters, Matt Harris, Kundu, and Chris Middleton. And you can find more information on today's episode in the show notes. Make sure to hit subscribe and give us a rating in your podcast app.
And join us next time when we'll be looking at containers: the what, the why, and the how [of] this small but mighty piece of technology.
Today's show was written and produced by Isobel Pollard and was hosted by me, Michael Bird, with sound design and editing by Alex Bennett and production support from Harry Morton, Thomas Berry, and Alex Padmore.
Technology Untangled is a Lower Street Production for Hewlett Packard Enterprise in the U.K. and Ireland.
Thanks for tuning in and we'll see you next time.
Listen to other episodes:
Hyperconverged infrastructure—because of the simplicity of it, because of the building block nature of it—means the test and development cycles on applications can be faster.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.