HP’s cloud guy: Why we’re the enterprise cloud

Just 16 months ago, Leo Apotheker, the short-lived CEO of HP, proclaimed that HP would be a leader in cloud computing. With little to show in the way of HP cloud products or services, no one was quite sure what Apotheker was talking about.

Flash-forward to today, and HP has not only a real cloud strategy, but also a public cloud IaaS (infrastructure as a service) play, HP Cloud, which might eventually rival Amazon Web Services. It also has Zorawar “Biri” Singh, senior vice president and GM of HP Cloud Services, to keep HP Cloud and various other cloud services and solutions on track.

This is not Singh’s first cloud gig — prior to arriving at HP, he served as vice president of Cloud Computing for IBM. In his new position, Singh sees the opportunity to couple HP’s pure-play IaaS cloud with private HP clouds that will “integrate very naturally” with each other. In that hybrid approach, Singh relies on OpenStack, the open source cloud operating system, to provide the underpinning for HP’s public and private cloud offerings.

From the outset of our conversation with Singh last week, he made it clear that he felt the “real opportunity” for HP was in serving developers, so that’s where this edited version of the interview begins. InfoWorld executive editor Doug Dineley also participated.

Eric Knorr: What sort of developers do you feel will be inclined to adopt HP’s cloud offerings?

Biri Singh: We’re going after the enterprise developer, where there are a bunch of expectations about which production workloads are going to end up on the public cloud. We happen to think there will be tens of thousands use cases that are ultimately going to be driven by the need for a secure, SLA-driven, enterprise-class quality of service. Our focus is the enterprise developer, but also IT ops.

For production workloads enterprises may consider running, they want the scale, they want the advantage of cost efficiencies. They want the security. But most importantly, they want a vendor who understands what they’re about, who they’ve done business with, who understands the need for innovative services yet can balance out SLA, security, and customer service — and who provides choices in terms of being an open architecture, partnering with other stacks and not locking in customers.

If you look at infrastructure clouds, AWS (Amazon Web Services) is obviously at one end of the spectrum, but you also have a bunch of telcos that have tried this, and their notion of cloud is — well, stand up a bunch of VMs and let’s see what happens. I think that’s such a 2009 phenomena; we’ve passed that. What developers and enterprises are looking for, they’re saying: Give me a set of services. I want to be able to run workloads in a secure cloud environment. And I want the best tools, the most modern languages and frameworks to build those. We’re trying to address the needs of developers and IT ops with a particular emphasis on the kind of enterprise, production-grade workloads are going to be running.

Knorr: Obviously, dev and test has been one of the big uses of the public cloud from the very beginning. But you’re talking about dev, test, and deploy for the enterprise, which is something different. From what we’ve seen, for the enterprise, it’s been dev and test in the public cloud, but not deploy in the public cloud. You bring it back in-house and deploy it on your own servers. How are you going to succeed in deploying enterprise apps in your public cloud where others have failed?

Singh: Let’s talk about deploying specifically; then I’ll talk about PaaS (platform as a service) as well. In terms of the deploy model, everyone has a sandbox on AWS or is trying, but they very quickly realize that in order to get to production they’ve got to pull these environments back in.

One of two things happens. CIOs will say, well, I’m 70 percent, 80 percent virtualized, so why can’t I just do that here? And running a dev/test application in a cloud setting versus running it in-house on a highly virtualized setting, there are different things you have to manage. It becomes a very different experience and technically different for the apps guys or the IT ops guys to manage that.

The other thing that happens is people get spoiled by the ease of dev/test, whether on AWS, whether on HP Cloud or any environment, because you get out of the old provisioning and the many weeks of provisioning and all that. But now you can also impact the application and the interaction with the applications. So people get used to the ease of deployment.

That’s why I think ultimately private clouds and public clouds become sort of one and the same. There’s an on-ramp and a secure orchestration model to it. But that’s what CIOs want.

People say: Oh, virtualization and cloud are really one thing. They’re very different, actually. With virtualization, you’re managing clusters of VMs and there’s a certain management and security model for that. You’re essentially building on your hypervisor, then most of the offerings drive you in sort of a lock-in mode.

Knorr: And the cloud?

Singh: It’s very different. You’re dealing with availability zones across network models whether flat or virtualized. You’re dealing with data redundancies and deployment models that have to do with underlying application architectures. It’s a very different experience doing dev/test and then dev/test/deploy. Our goal is to give enterprises the complete lifecycle.

Now, people have an option to use different PaaS environments to deploy as well. But what enterprises are very quickly arriving at is — if they build a Ruby app on Engine Yard or a PHP app on AppFog or something on Jenkins driven with CloudBees — they want to understand long term how those apps end up running in their environment. And they don’t want to do the heavy lifting of provisioning those cloud-like environments. They just want to be able to build and focus on running the business process or the app that they’ve rewritten in a PaaS environment, and they just want it hosted.

If you look at our ecosystem play, we’ve built out a whole bunch of partnerships with over 80 companies that allow you to do things like PaaS deployments. Our view on PaaS is we want to offer choice.

But it comes down to a very simple premise: The CIO says, hey, it’s great that you’re building these new-wave mobile apps and these new Ruby apps or whatever, and rewriting traditional boring legacy stuff, but where are all these apps going to run and how do we have control over them? Where’s the SLA? Where’s the security layer? Where’s the isolation layer? Where’s my disaster recovery? If you’re trusting Engine Yard or you’re trusting CloudBees or you’re trusting whoever to run those, how do we ultimately instantiate that for our needs? Ultimately, large-scale public clouds matter, SLAs matter, so we’re working with PaaS developers — not to be the end-all PaaS. HP has no interest in being a kingmaker of Ruby or Python or PHP or whatever.

Knorr: Is that part of the idea behind adopting OpenStack, to offer enterprises this kind of a parallel hybrid environment?

Singh: We’re talking about two different levels of HP Cloud here. One is the HP Converged Cloud, which is our umbrella strategy for hybrid delivery of private cloud, managed cloud services, and public cloud. This is the notion of hybrid delivery for enterprises, for service providers, for customers across different environments.

We have private cloud solutions as a product, CloudSystem. We have managed cloud as a consulting services-driven model where we do CloudSystem as a service, essentially private cloud hosting, but also public cloud instantiation. It’s where we go to our outsource customers and say, hey, if you want a bare-metal environment and just run Red Hat or Windows or Linux, I could turn that on for you. I can manage your applications for you. I can do application transformation for you. I can host Exchange, rewrite your CRM app, JVM app, etc.

Knorr: Certainly, that’s not something you would fire up with a credit and a Web form on your own.

Singh: No, that’s what the public cloud is for. You can go to HPcloud.com/partners and actually see our cloud ecosystem at work. We’re working with folks like Active State and AppFog and Engine Yard and AppZero and CloudBees or on and on. We also offer storage and management and monitoring and data services and dev/test services. If I’m going to host a bunch of PaaS environments on the HP Cloud, it’s going to be reasonably automated — it’s not a consulting services deployment. You can go on and actually turn on and fire up through a command line or through our management console. You can use a bunch of tools and offerings we’ve built up through ecosystem and third-party partners.

Doug Dineley: Is it fair to say that all PaaS offerings on the HP Cloud will be through partners, or is HP going to provide some of its own PaaS services, such as relational database as a service?

Singh: Let’s break down platform services as follows. There is a notion of PaaS in terms of development tools and IDEs. We’re working with partners there. Eventually, HP will also have an offering — but we’re not trying to control languages and frameworks. We think our job is to host as many in a meaningful way and give our developers and our IT ops folks choices.

The second part of it is doing things like database as a service. We’ve launched MySQL as a service, which our team has built and is a really important offering. And I think, if you really boil it down … the value prop of what a successful PaaS offering should be for a developer or IT ops folks … all they really care about is their runtime and the elastic database behind it. But they also want everything else, like autoscale, load balancing, security, identity management, billing, and so on. So we’re doing that.

The third thing is analytics as a service. The ability for developers and IT ops folks to very quickly and in real time tie in all the work and the workloads that they’re building and do real-time analysis of that. Then when you add in things like management, monitoring, security as a service, billing as a service, storage as a service, all of a sudden the platform that we’re talking about starts becoming a much more vibrant offering.

Knorr: Where does OpenStack come in?

Singh: With OpenStack, we’re using Nova, Swift, Quantum — we’ve done a heavy curation of it, not a distribution. We did it on the Diablo code stack. Our next big curation release will be on Folsom [a more complete version of OpenStack due this fall]. So we’re running our public cloud today, it’s in thousands of nodes, it’s one of the largest OpenStack implementations out there, and we’re going to go pretty heavy with Folsom. But I’m also building a private cloud reference architecture that I’ve integrated on top of HP’s service storage networking, and we’re going to have a completely integrated stack.

Knorr: Would you characterize those as proprietary extensions to OpenStack?

Singh: No, I would not. I’d characterize those as OpenStack API, EC2-compatible, API-driven extensions of HP’s hardware service storage networking running the OpenStack distribution that HP has curated. We are opening up our APIs for everyone to basically subscribe to. We are going to support the OpenStack API, as well as EC2, to again give choice and there’s nothing “proprietary” about it.

Knorr: When you use words like “curated,” are you saying what you’re doing is based on the OpenStack bits themselves, or is it really mostly about API compatibility?

Singh: It is actually built on the OpenStack bits. We’ve added a bunch of things on top of that. For example, we’ve done a whole bunch of work on the Nova compute platform as part of OpenStack. The Keystone service of OpenStack (which is essentially sort of the control services layer, you do things like identity management, security, billing, metering integration), we’ve done a whole bunch of work there to meet the needs of the enterprise and service providers.

Knorr: OpenStack has its own trajectory as it hammers out specifications. There have also been some defections within OpenStack, such as the departure of Citrix. And Nicira, which was central to developing Quantum, was purchased by VMware, which has not been a robust supporter of OpenStack ….

Singh: Got it. I understand where you’re going with this.

First things first. OpenStack has a pretty large community. There are some large industry heavyweights — you know, HP, AT&T, IBM, Red Hat, Cisco, Rackspace, Dell — so I don’t think OpenStack’s going away. It has thousands of downloads, thousands of participants, a large community. We’re an active member. We don’t see it slowing down anytime soon.

You could say that VMware acquiring Nicira is a shot at Quantum, but I would counter that. If you actually talk to Nicira … they are very adamant, they’re tweeting this morning their support for OpenStack going forward. So when you say OpenStack has its own trajectory, with all due respect, it’s a trajectory that’s informed by the likes of HP, the likes of Red Hat, IBM, AT&T. We’re certainly a big player in both the public and private part of that.

Knorr: Are you a big contributor of code to OpenStack?

Singh: We are the sixth-largest contributor — the second-largest contributor in Swift and one of the top three contributors in Nova, Quantum, Keystone, and Glance. Go look at the code base. We’re there. And if you look at the community forums, we’re pretty active.

By the way, Rackspace, Red Hat, and Nebula have been there, too. You know, Red Hat put in a bunch of code around their model. Code speaks, and I’d invite you to go look at it.

I don’t think OpenStack’s going anywhere. Having said that, it’s still early days, but we intend to absolutely drive the next-generation computer infrastructure, and we think OpenStack is an important ingredient. It’s not the only one, but it’s certainly a very important part of it.

Dineley: Have all your improvements to OpenStack for HP Cloud been contributed back to the project?

Singh: Yeah, by Folsom, you’re going to see meaningful interactions. I think only literally a handful of people on the planet really know what they’re talking about. We are one of them. And please don’t take that as an arrogant thing. Community code is not the same as production code when you get to large-scale cloud deployment in the thousands and tens of thousands of nodes. You cannot take community code and go publish that. So we’re going to contribute back; we already are.

Our design point is to run the HP Cloud for all the value properties I talked about earlier. I don’t think that equates to saying: Oh, yeah, you can just take the latest Folsom and go stand up a 10,000-node cluster. It’s not going to happen. There are a lot of unique things you have to manage in order to do that. That’s where the breadth and depth of HP comes into play.

This article, “HP’s cloud guy: Why we’re the enterprise cloud,” originally appeared at InfoWorld.com. Read more of Eric Knorr’s Modernizing IT blog. And for the latest business technology news, follow InfoWorld on Twitter.