Interview conducted by Vinod Baya, Bo Parker, and Victoria Huff
Russ Daniels is CTO of HP Enterprise Services. In this role, announced in June 2009, he continues the momentum he built as vice president and Cloud Services CTO at HP, driving the cloud technology strategy and architecture for HP’s outsourcing services portfolio. Daniels has more than 25 years of experience in the technology industry, specializing in software architecture, enterprise management, and software development methodologies. He has filled a wide range of staff and line management roles and run his own Internet services business. From 2002 to 2007, Daniels was CTO of Software for the Technology Solutions Group at HP. During his tenure, the business tripled in revenue and emerged as a significant player in the software industry. In 2006, InfoWorld named Daniels one of the industry’s top 25 CTOs.
In this interview, Daniels shares insights on how IT will create agility by moving from a predominantly build-to-order model to a configure-to-order model. He also discusses how cloud computing represents an opportunity to create new value at the level of industry ecosystems.
PwC: What level of disruption does cloud computing represent? Does it represent the same level as the Internet, something smaller, or even something bigger?
RD: We think two things are going on. The first primarily concerns what a year ago we would have called utility computing rather than cloud computing: the fundamental questions of how do you improve the efficiencies and the agility of the IT function as it sources and delivers the services the business needs. It’s a very important agenda, but it isn’t new. It provides some new opportunities, some new choices that businesses can make about how they source and deliver those services. We think calling that cloud, though, is confusing, because it obscures or takes away the ability to talk about the second thing that’s going on that we think has more long-term and transformational implications.
A 2x2 diagram is useful to show how we see things. [See Figure 1.] The left column describes on-premises, which is to say the compute capacity, the data center, the servers, the storage, the network capacity, the software licenses, and the operational responsibilities live within the enterprise, within the consumer—the consumer and producer are the same. Off-premises means some third party takes over more of those production responsibilities.
The bottom row is labeled “Dedicated,” and the one above it is labeled “Shared.” The idea is that in the “dedicated” row, the resources that are being applied are bound to specific application workloads, and making changes to those bindings requires manual intervention. In the “shared” row, there’s some mechanism that allows you to bind those resources more flexibly and to adjust those bindings dynamically.
PwC: What do each of the quadrants represent in this framing?
RD: If I walk you through the quadrants, the quadrant labeled 1 is the traditional way we built things in the past, which is the customer buys all the assets, constructs a complex execution container for a particular workload, and then effectively pours that container into cement—it’s very difficult to change. That approach is really good for some things. It’s great if you have concerns around performance, for example, because you can optimize everything in that environment for performance, availability, and throughput. So, for things that are mission critical—which is to say something such that when it goes down, your business can’t book revenue, can’t produce products, services, et cetera—you might really like this model, because it provides you with the ultimate control.
If you need some of those characteristics, but you don’t have the competencies or don’t want to maintain the competencies to do that, you might move to the right to quadrant 2, where somebody else does some of that stuff for you. For most of our customers, when they think of the cloud, they’re thinking of moving from quadrant 1 to 3, where they typically take advantage of some combination of virtualization and automation technologies. They can pool the resources and then dynamically bind them to workloads and adjust those bindings as business need dictates. The upper right quadrant, quadrant 4, is where you take advantage of shared capabilities off-premises, so some third party offers the services that the business is consuming. Here you get the advantages of scaled multitenancy and potentially specialization to be able to do particular things at a cost efficiency that makes it attractive.
PwC: How does the framework in the diagram relate to the responsiveness of IT to changes?
RD: The best way to answer that is to change these labels from dedicated and shared, and say what’s really going on in the context of this frame. If you use the manufacturing metaphor, the bottom world is the world of build to order [BTO]. Build to order is distinct in that design is an integral aspect of delivery. It’s custom. It’s bespoke tailoring. You want a suit? No problem. Here are some bolts of cloth. Which ones do you like? What felt would you like under the collar? What buttons would you like? We will create exactly the suit you want. It’s going to be a little bit more expensive than if you go and buy it off the shelf. The delivery might be six weeks from now rather than later today—up to you.
The world in the top row is the world of configure to order [CTO], and configure to order is a world that—again, in the tailoring world, it means you buy off the rack. There’s a set of designs. You might be able to get a little bit of alteration done, but if you’re a size 44 jacket and a size 42 pants, then you’re going to have to find the right shop to buy from. So there are all these different things that you give up, but you get lower cost and faster availability.
So, for the IT function, the opportunities and the significant impacts are about shifting your culture from a build-to-order world, where everything is done in a bespoke model, to a configure-to-order world. This is just as relevant within an IT organization as it is for a service provider. For example, EDS has a lot of culture and competencies around build to order, and a lot of the opportunities that we can address from a business perspective concern how we capture that best-practice knowledge in a form that allows us then to apply it in a configure-to-order delivery model.
PwC: How does your framework apply to cloud computing?
RD: We don’t think any of this has anything to do with the cloud other than one important dependency. We think quadrant 1 is the world of classic IT. Quadrant 3 is the world of utility computing. It is what HP talks about as adaptive infrastructure, what Gartner calls real-time infrastructure, what Forrester calls organic IT. There’s a huge amount of pressure in the industry to grab onto a new term and to recharacterize everything we already do as embodying the values of that new term. So, people all over the industry have taken their existing slide decks and slapped “cloud” around liberally, primarily for the things that are here in quadrants 3 and 4.
It’s hard not to call it cloud. I’m just saying that it doesn’t really clarify anything. But there is something else going on, and let’s call that cloud. If you look at the figure again, you’ll see I’ve added above the four existing quadrants a private cloud and a public cloud.
And so the key question to ask is what was added to our four existing quadrants to justify a shift from just talking about utility to talking about cloud? Our answer is that it’s about how you design. It’s primarily around software, but you can think about software and systems. You get fundamentally different results and expressiveness, allowing you to solve a different class of problems—problems that when you try to solve them in traditional delivery models, the projects tend to be either so expensive they never get started or so complex that they underdeliver.
PwC: To move from BTO to CTO, what is the role of automation of IT processes? What are the unsolved challenges in automation to achieve a configure-to-order future?
RD: We think automation ultimately ends up being more important, because you can automate both physical as well as virtual environments and support configure to order of more and more complex application architectures. Additionally, if you have the ability to rethink the way you architect workloads, it then allows you to attack a whole set of problems that you can never address in the context of traditional application designs.
Another thing about automation is that the traditional approaches to automation are significantly focused on imperative logic. So [for provisioning], either it’s scripts, and scripts are software with a particular notation; or it’s workflows, which are software with a particular notation—it’s all imperative logic.
And our view is that to successfully automate more and more complex kinds of configure-to-order deliveries, you really need to focus on forms that are more declarative rather than imperative in their expression. We have a strong focus on how to capture the relevant information in information models and then allow systems software to perform introspections on those models and to understand what must happen in a sequential workflow-centric approach.
The advantage of that information model is that you can leverage it in multiple ways. It also ends up being just as simple. It’s much easier to generate data than it is to generate code.
PwC: HP has been supporting the management of IT environments via its OpenView platform. Where are we with managing the infrastructure in cloud environments?
RD: In many cases, the key to understanding the transformation that’s going on right now is to realize that most of the management technologies in the market today have focused on addressing the needs that come from this build-to-order world—complex heterogeneous environments with a huge degree of customization. Another key component is that in the BTO world, there’s a very strong distinction between the creation of the service versus the ongoing delivery and operation of the service. The classic model is that the software developer builds something and then throws it over the wall to the operations team who then has to figure out how to make it work well enough to meet the business needs. Because of the convergence that we see between those development and deployment activities [in CTO], you then have the ability to shift more of the responsibility back in the pre-production concerns.
So if you look at our management portfolio, it isn’t just about operations management. When we brought in Mercury, we brought in a lot of capabilities related to quality management, performance testing, and validation. “How can I do more work earlier in the life cycle to characterize a workload to be able to take out risk in operations?” And the key thing that happens then is you turn down some of the pressure on improved process maturity of IT organizations and, instead, push it into, “Let’s try to get more of these things addressed architecturally.”
PwC: What will be the impact of cloud computing on enterprises? What is the new opportunity for them?
RD: During the last 20 years, there’s been a lot of increased growth and interest around enterprise architecture, where enterprise architecture really focuses on understanding the behavior of the enterprise itself and where the organizing abstraction is the business process. So the highest-level organizing structure within enterprise architecture comprises those fundamental business processes, order to cash, and so on.
We think the cloud allows you to step up to yet another level of architectural design concern, where, rather than thinking about the processes the business uses to accomplish its end, you think about the business in the context of the ecosystems in which it creates value. So it’s the business in context. How do you interact with your customers? How do you interact with your channels, your partners, your suppliers? It’s those abilities to understand, to share information, to encourage the useful collaborations that you need within those ecosystems—that’s why we think the cloud is so important.
PwC: Does that need to be in a shared cloud, ideally, and therefore something that is a result of public cloud services?
RD: We think the cloud ultimately is all about using information and focusing on sharing information with the intended purpose of improving collaboration across an ecosystem. And, again, there’s this subtle distinction that within an ecosystem you can’t require conformity around a business process.
It’s those ecosystem interactions, and this comes down to our definition of the cloud, which is that it’s the next phase of evolution of the Internet. The characteristic of the next phase of the Internet is the ability to capture huge amounts of data—adding and extending, and creating more and more context. That characteristic enables you to create much richer kinds of interactions in the context of ecosystems that you care about.
PwC: Can you provide an example of how cloud architecture will enable new interactions and value in an ecosystem?
RD: Let me give you one practical example. Think about the issues of product traceability in the produce supply chain. Was the salmonella outbreak because of the tomatoes or the jalapeños, or maybe it was the onions or the cilantro—because it was found in salsa, right? We figured that out. If you try to think about how you would answer that kind of question from a technical implementation perspective in our traditional architectures, you would need point-to-point integrations between the supply chain management systems of all the participants in that extended supply chain. That’s a very expensive and complex thing to do, and the value of being able to answer these questions isn’t great enough that you’ll go through that cost and complexity. So when we try to answer them today, practically speaking, people are pouring through printed reports to try to figure out what happened.
You’ve minimized the complexity for all the participants. There was no point-to-point integration. All you had to do was generate a report, which is an easy thing to do, and if you add a new player in the supply chain, it’s just one more report that gets generated. Somebody leaves, and that report’s no longer generated. You don’t need to do anything more at the nodes than that. All the cost and the complexity of the analysis goes to the cloud service itself.