Bringing higher levels of automation to data center infrastructure has long been a priority for IT operators, but it's only been in the past few years that they have actually enjoyed truly workable solutions for composability.
The growing complexities, from hybrid cloud and the pressing need for conservation of IT spend -- as well as the need to find high-level IT skills -- means there is no going back. Indeed, there is little time for even a plateau on innovation around composability.
Stay with us now as we explore how pervasive increasingly intelligent IT automation and composability can be with Gary Thome, Vice President and Chief Technology Officer for Composable Cloud at Hewlett Packard Enterprise (HPE). The interview is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.
Here are some excerpts:
Gardner: Gary, what are the top drivers making composability top-of-mind and something we’re going to see more of?
Thome: It’s the same drivers for businesses as a whole, and certainly for IT. First, almost every business is going through some sort of digital transformation. And that digital transformation is really about transforming to leverage IT to connect with their customers and make IT the primary way they interact with customers and make revenue.
Digital transformation drives composability
With that, there’s a desire to go very fast, of rapidly getting connections to customers much faster and for adding features faster via software for your customers.
The whole idea of digital transformation and becoming a digital business is driving a whole new set of behaviors in the way enterprises run – and as a result – in the way that IT needs to support them.
From the IT standpoint, there is this huge driver to say, “Okay, I need to be able to go faster to keep up with the speed of the business.” That is a huge motivator.
But at the same time, there’s the constant desire to keep IT cost in line, which requires higher levels of automation. That automation -- along with a desire to flexibly align with the needs of the business -- drives what we call composability. It combines the flexibility of being able to configure and choose what you need to meet the business needs -- and ultimately customer needs -- and do it in a highly automated manner.
Gardner: Has the adoption of cloud computing models changed the understanding of how innovation takes place in an IT organization? There used to be long periods between upgrades or a new revision. Cloud has given us constant iterative improvements. Does composability help support that in more ways?
Thome: Yes, it does. There has been a general change in the way of thinking, of shifting from occasional, large changes to frequent, smaller changes. This came out of an Agile mindset and a DevOps environment. Interestingly enough, it’s permeated to lots of other places outside of IT. More companies are looking at how to behave that way in general.
How to Achieve Composability
Across Your Datacenter
On the technology side, the desire for rapid, smaller changes means a need for higher levels of automation. It means automating the changes to the next desired state as quickly as possible. All of those things lend themselves toward composability.
Gardner: At the same time, businesses are seeking economic benefits via reducing unutilized IT capacity. It’s become about “fit-for-purpose” and “minimum viable” infrastructure. Does composability fit into that, making an economic efficiency play?
Thome: Absolutely. Along with the small, iterative changes – of changing just what you need when you need it – comes a new mindset with how you approach capacity. Rather than buying massive amounts of capacity in bulk and then consuming it over time, you use capacity as you need it. No longer are there large amounts of stranded capacity.
Composability is key to this because it allows you through technical means to gain an environment that gets the desired economic result. You are simply using what you need when you need it, and then releasing it when it’s not needed -- versus pre-purchasing large amounts of capacity upfront.
Innovation building blocks
Gardner: As an innovator yourself, Gary, you must have had to rethink a lot of foundational premises when it comes to designing these systems. How did you change your thinking as an innovator to create new systems that accommodate these new and difficult requirements?
Thome: Anyone in an innovation role has to always challenge their own thinking, and say, “Okay, how do I want to think differently about this?” You can't necessarily look to the normal sources for inspiration because that's exactly where you don't want to be. You want to be somewhere else.
For myself it may mean looking at any other walk of life – from what I do, read, and learn as possible sources of inspiration for rethinking the problem.
Interestingly enough, there is a parallel in the IT world of taking applications and decomposing them into smaller chunks. We talk about microservices that can be quickly assembled into larger applications -- or composed, if you want to think of it that way. And now we’re able to disaggregate the infrastructure into elements, too, and then rapidly compose them into what's needed.
Those are really parallel ideas, going after the same goal. How do I just use what I need when I need it -- not more, not less? And then automate the connections between all of those services.
That, in turn, requires an interface that makes it very easy to assemble and disassemble things together -- and therefore very easy to produce the results you want.
When you look at things outside of the IT world, you can see patterns of it being easy to assemble and disassemble things, like children's building blocks. Before, IT tended to be too complex. How do you make the IT building blocks easier to assemble and disassemble such that it can be done more rapidly and more reliably?
Gardner: It sounds as if innovations from 30 years ago are finding their way into IT. Things such as simultaneous engineering, fit-for-purpose design and manufacturing, even sustainability issues of using no more than you need. Were any of those inspirations to you?
Cultivate the Agile mindset
Thome: There are a variety of sources, everything from engineering practices, to art, to business practices. They all start swiveling around in your head. How do I look at the patterns in other places and say, “Is that the right kind of pattern that we need to apply to an IT problem or not?”
The historical IT perspective of elongated steps and long development cycles led to the end-place of very complex integrations to get all the piece-parts put together. Now, the different, Agile mindset says, “Why don’t you create what you need iteratively but make sure it integrates together rapidly?”
Can you imagine trying to write a symphony and have 20 different people develop their own parts? There’s separate trombone, or timpani, or violin. And then you just say, “Okay, play it together once, and we will start debugging when it doesn’t sound right.” Well, of course that would be a disaster. If you don’t think about it upfront, do you want to develop it as-you-go?
The same thing needs to go into how we develop IT -- with both the infrastructure and applications. That’s where the Agile and the DevOps mindsets have evolved to. It’s also very much the mindset we have in how we develop composability within HPE.
Gardner: At HPE, you began bringing composability to servers and the data center stack, trying to make hardware behave more like software, essentially. But it’s grown past that. Could you give us a level-set of where we are right now when it comes to the capability to compose the support for doing digital business?
Intelligent, rapid, template-driven assembly
Thome: Within the general category of composablity, we have this new thing called Composable Infrastructure, and we have a product called HPE Synergy. Rather than treat the physical data resources in the data center as discrete servers, storage arrays, switches, it looks at them as pools of compute capacity, storage capacity, fabric capacity, and even software capacity or images of what you want to use.
Each of those things can be assembled rapidly through what we call software-defined intelligence. It knows how to assemble the building blocks – compute, storage, and networking -- into something interesting. And that is template-driven. You have a template, which is a description of what you want the end-state to look like, what you want your infrastructure look like, when you are done.
And the templates say, “Well, I need a compute of this big block or size. This much storage, or this kind of network.” Whatever you want. “And then, by the way, I want this software loaded on it.” And so forth. You describe the whole thing as a template and then we can assemble it based on that description.
That approach is one we’ve innovated on in a lab from the infrastructure’s standpoint. But what’s very interesting about it is, if you look at a modern cloud made up of applications, it uses a very similar philosophical approach to the assembling. In fact, just like with modern applications, you say, “Well, I’m assembling a group of services or elements. I am going to create it all via APIs.” Well, guess what? Our hardware is driven by APIs also. It’s an API-level assembly of the hardware to compose the hardware into whatever you want. It’s the same idea of composing that applies everywhere.
Millennials lead the way
Gardner: The timing for this is auspicious on many levels. Just as you’re making crafting of hardware solutions possible, we’re dealing with an IT labor shortage. If, like many Millennials, you are of a cloud-first mentality you will find kinship with composability -- even though you’re not necessarily composing a cloud. Is that right?
Thome: Absolutely. That cloud mindset, or service’s mindset, or asset-service mindset -- whatever you want to think of it as – is one where this is a natural way of thinking. The younger people may have grown up with this mindset. It wouldn’t occur to them to think any differently. And others may have to shift to a new way of thinking.
This is one of the challenges for organizations. How do they shift -- not just the technologies or the tools -- but the mindset within the culture in a different direction?
How to Remove Complexity
From Multicloud and Hybrid IT
You have to start with changing the way you think. It’s a mindset change to ask, “How do I think about this problem differently?” That’s the key first thing that needs to happen, and then everything falls behind that mindset.
It’s a challenge for any company doing transformation, but it’s also true for innovation -- shifting the mindset.
Gardner: The wide applicability of composability is impressive. You could take this composable mindset, use these methods and tools, and you could compose a bare-metal, traditional, on-premises data center. You could compose a highly virtualized on-premises data center. You could compose a hybrid cloud, where you take advantage of private cloud and public cloud resources. You can compose across multiple types of private and public clouds.
Thome: We think composability is a very broad, useful idea. When we talk to customers they are like, “Okay, well, I’ll have my own kind of legacy estate, my legacy applications. Then I have my new applications, and new way of thinking that are being developed. How do I apply principles and technologies that are universal across them?”
The idea of being able to say, “Well, I can compose the infrastructure for my legacy apps and also compose my new cloud-native apps, and I get the right infrastructure underneath.” That is a very appealing idea.
But we also take the same ideas of composability and say, “Well, I would even want to compose ultimately across multiple clouds.” So more-and-more enterprises are leveraging clouds in various shapes and forms. They are increasing the number of clouds they use. We are trending to hybrid cloud, where there are people using different clouds for different reasons. They may actually have a single application that’s spanning multiple clouds, including on-premises clouds.
When you get to that level, you start thinking, “Well, how do I compose my environment or my applications across all of those areas?” Not everybody is necessarily thinking about it that way yet, but we certainly are. It’s definitely something that’s coming.
You start thinking, "How do I compose my environment or my applications across all areas?" Not everyone is thinking about it yet that way, but we certainly are. It's definitely coming.
Gardner: Providers are telling people that they can find automation and simplicity but the quid pro quo is that you have to do it all within a single stack, or you have to line up behind one particular technology or framework. Or, you have to put it all into one particular public cloud.
It seems to me that you may want to keep all of your options open and be future-proof in terms of what might be coming in a couple of years. What is it about composability that helps keep one’s options open?
Thome: With automation, there’s two extremes that people wind up with. One is a great automation framework that promises you can automate anything. The most important thing is that you can; meaning, wedon’t do it, but youcan, if you are willing to invest all of the hard work into it. That’s one approach. The good news is that there are multiple vendors with actual parts of the automation-technology total. But it can be a very large amount of work to develop and maintain systems across that kind of environment.
On the other hand, there are automation environments where, “Hey, it works great. It’s really simple. Oh, by the way, you have to completely stay within our environment.” And so you are stuck within the confines of their rules for doing things.
Both of these approaches, obviously, have a very significant downside because any one particular environment is not going to be the sum of everything that you do as a business. We see both of them as wrong.
Real composability shines when it spans the best of both of those extremes. On the one hand, composability makes it very easy to automate the composable infrastructure, and it also automates everything within it.
In the case of HPE Synergy, composable management (HPE OneView) makes it easy to automate the compute, storage, and networking -- and even the software stacks that run on it -- through a trivial interface. And at the same time, you want to integrate into the broader, multivendor automation environments so you can automate across all things.
You need that because, guaranteed, no one vendor is going to provide everything you want, which is the failing of the second approach I mentioned. Instead, what you want is to have a very easy way to integrate into all of those automation environments and automation frameworks without throwing a whole lot of work to the customer to do.
We see composability strength in being API-driven. It makes it easy to integrate into automation frameworks, but secondly, it completely automates the things that are underneath that composable environment. You don't have to do a lot of work to get things operating.
So we see that as the best of those two extremes that have historically been pushed on the market by various vendors.
Gardner: Gary, you have innovated and created broad composability. In a market full of other innovators, have there been surprises in what people have done with composability? Has there been follow-on innovation in how people use composability that is worth mentioning and was impressive to you?
Thome: One of my goals for composability was that, in the end, people would use it in ways I never imagined. I figured, “If you do it right, if you create a great idea and a great toolset, then people can do things with it you can't imagine.” That was the exciting thing for me.
One customer created an environment where they used the HPE composable API in the Terraform environment. They were able to rapidly span a variety of different environments based on self-service mechanisms. Their scientist users actually created the IT environments they needed nearly instantly.
It was cool because it was not something that we set out specifically to do. Yet they were saying it solves business needs and their researchers’ needs in a very rapid manner.
Another customer recently said, “Well, we just need to roll out really large virtualization clusters.” In their case, it's a 36-node cluster. It used to take them 21 days. But when they shifted to HPE composability, they got it down to just six hours.
How to Tame
Obviously it’s very exciting to see such real benefits to customers, to get faster with putting IT resources to use and to minimize the burden on the people associated with getting things done.
When I hear those kinds of stories come back from customers -- directly or through other people -- it's really exciting. It says that we are bringing real value to people to help them solve both their IT needs and their business needs.
Gardner: You know you’re doing composable right when you have non-IT people able to create the environments they need to support their requirements, their apps, and their data. That's really impressive.
Gary, what else did you learn in the field from how people are employing composability? Any insights that you could share?
Thome: It's in varying degrees. Some people get very creative in doing things that we never dreamed of. For others, the mindset shift can be challenging, and they are just not ready to shift to a different way of thinking, for whatever reasons.
Gardner: Is it possible to consume composability in different ways? Can you buy into this at a tactical level and a strategic level?
Thome: That's one of the beautiful things about the HPE composability approach. The answer is absolutely, “Yes.” You can start by saying, “I’m going to use composability to do what I always did before.” And the great news is it's easier than what you had done before. We built it with the idea of assembling things together very easily. That's exactly what you needed.
Then, maybe later, some of the more creative things that you may want to do with composability come to mind. The great news is it's a way to get started, even if you haven’t yet shifted your thinking. It still gives you a platform to grow from should you need to in the future.
Gardner: We have often seen that those proof-points tactically can start the process to change peoples' mindsets, which allows for larger, strategic value to come about.
Thome: Absolutely. Exactly right. Yes.
Gardner: There’s also now at HPE, and with others, a shift in thinking about how to buy and pay for IT. The older ways of IT, with longer revisions and forklift upgrades meant paying was capital-intensive.
What is it about the new IT economics, such as HPE GreenLake Flex Capacity purchasing, that align well with composability in terms of making it predictable and able to spread out costs as operating expenses?
Thome: These two approaches are perfect together; they really are. They are hand-in-glove and best buddies. You can move to the new mindset of, “Let me just use what I need and then stop using it when I don't need it.”
That mindset -- and being able to do rapid, small changes in capacity or code or whatever you are doing, it doesn’t matter – also allows a new economic perspective. And that is, “I only pay for what I need, when I need it; and I don't pay for the things I am not using.”
Our HPE GreenLake Flex Capacity service brings that mindset to the economic side as well. We see many customers choose composability technology and then marry it with GreenLake Flex Capacity as the economic model. They can bring together that mindset of making minor changes when needed, and only consuming what is needed, to both the technical and the economic side.
We see this as a very compelling and complementary set of capabilities -- and our customers do as well.
Gardner: We are also mindful nowadays, Gary, about the edge computing and the Internet of Things (IoT), with more data points and more sensors. We also are thinking about how to make better architectural decisions about edge-to-core relationships. How do we position the right amount of workload in the right place for the right requirements?
How does composability fit into the edge? Can there also be an intelligent fabric network impact here? Unpack for us how the edge and the intelligent network foster more composability.
Composability on the fly, give it a try
Thome: I will start with the fabric. So the fabric wants to be composable. From a technology side, you want a fabric that allows you to say, “Okay, I want to very dynamically and easily assemble the network connections I want and the bandwidth I want between two endpoints -- when I want them. And then I want to reconfigure or compose, if you will, on the fly.”
We have put this technology together, and we call it Composable Fabric. I find this superexciting because you can create a mesh simply by connecting the endpoints together. After that, you can reconfigure it on the fly, and the network meets the needs of the applications the instant you need them.
How to Achieve Composability
Across Your Datacenter
This is the ultimate of composability, brought to the network. It also simplifies the management operation of the network because it is completely driven by the need from the application. That is what directly drives and controls the behavior of the network, rather than having a long list of complex changes that need to be implemented in the network. That tends to be cumbersome and winds up being unresponsive to the real needs of the business. Those changes take too long. This is completely driven from the needs of application down into the needs of the fabric. It’s a super exciting idea, and we are really big on it, obviously.
Now, the edge is also interesting because we have been talking about conserving resources. There are even fewer resources at the edge, so conservation can be even more important. You only want to use what you need, when you need it. Being able to make those changes incrementally, when you need them, is the same idea as the composability we have been talking about. It applies to the edge as well. We see the edge as ultimately an important part of what we do from a composable standpoint.
Gardner: For those folks interested in exploring more about composability, methodologies, technologies, and getting some APIs to experiment with -- what advise do you have for them? What are some good ways to unpack this and move into a proof-of-concept project?
Thome: We have a lot of information on our website, obviously, about composability. There is a lot you can read up on, and we encourage anybody to learn about composability through those materials.
They can also try composability because it is completely software-defined and API-driven. You can go in and play with the composable concepts through software. We suggest people try directly. But they can also go and connect it to their automation tools and see how they can compose things via the automation tools they might already be using for other purposes. It can then extend into all things composable as well.
I definitely encourage people to learn more, but specially to move into the “doing phase.” Just try it out, see how easy it is to get things done.