Using virtualization to break vertical integration in IT silos

Photo: Simon Crosby of CitrixSimon Crosby of Citrix explains how virtualization creates separation in the IT vertical stack.

Interview conducted by Vinod Baya and Bo Parker

Simon Crosby is the CTO of the Virtualization and Management Division of Citrix Systems, and he was the founder and CTO of XenSource prior to the acquisition of XenSource by Citrix Systems. Prior to XenSource, Crosby was a principal engineer at Intel where he led strategic research in distributed autonomic computing, platform security, and trust. Previously, he was the founder of CPlane Inc., a network optimization software vendor. Before joining the private sector, Crosby was a tenured faculty member at the University of Cambridge, UK, where he led research on network performance and control, and multimedia operating systems. He is the author of more than 35 research papers and patents on data center and networking topics, including security, network and server virtualization, resource optimization, and performance.

In this interview, Crosby shares how the vertically integrated stacks in IT are a key barrier to agility and how virtualization technology is the software layer that will allow enterprises to bridge IT silos at the appropriate level based on their core competence.

PwC: Cloud computing is getting a lot of attention these days. How much disruption does it represent? Is it of the same scale as the Internet?

SC: It’s a profound change of the same order as the Internet—maybe somewhat smaller in the sense that today, cloud is pretty much articulated in the context of enterprise uses. But it’s of the same order in the sense that the Internet built the pipes to everybody—consumer and enterprise—and the cloud can deliver the apps to anybody—consumer or enterprise. It’s a different way to deliver all the applications that we consume today. I think it’s going to take a long time to get there, but it is a built-in process of maturing IT along the lines of maturing any utility function, so it’s about IT maturing into a planetary-wide utility.

PwC: We’re interested in the role IT infrastructure plays in enabling business agility. What are some of the key barriers to IT being responsive to businesses today?

SC: Today’s IT infrastructures are built from the equipment of today or 10 years ago and the successes of a scaled-out environment, which was entirely dependent on a growing economy. Given the dynamics and the fast pace of business, throwing in a new server and all the gunk that ran on it, and racking and stacking, was easier than addressing the fundamental management challenges of a scaled-out environment.

Now enterprises find themselves with these messily scaled infrastructures and an incapacity to deal with the ensuing complexity. Wherever the technology doesn’t provide the right degree of scale or visibility or anything else, you build an organization of humans around it to make sure that it runs. You end up with an organization that has a practiced excellence at handling servers, storage, the network, the various functions in the network, security, compliance, and all of these different things in the enterprise.

Additionally, vertically integrated groups understand how to make firewalls work, say, or how to make servers rack and stack and show up in data centers. But unfortunately, that vertical stack of knowledge about how to make a particular function work in the IT infrastructure is a direct enemy of deploying a new thing, because new things change every one of those silos, and to get a new thing deployed is going to require the IT team to change how a firewall works or the configuration of a particular set of devices and so on. Every one of the silos gets hit by a new deployment.

The enemy of agility is the existing vertical integration of expertise and function in the organization structures of IT. Even if you could get somebody to put a new VM [virtual machine] on the server, a network person will need to configure the VLANs [virtual local area networks] so all the traffic will show up, or configure the firewall purposely to let the traffic through. The organizational structures that have evolved are incompatible with the desire to be agile, and that is nobody’s fault. It’s property of the fact that the technologies were not mature and not self-scaling. What cloud gives us is a very interesting opportunity to change that.

Moreover, the IT infrastructure of today’s enterprise is massively diverse. The multivendor nature of it also conspires to cause complexity. Previously, multivendor sourcing was the enterprise’s friend in terms of retaining cost or pricing advantage from multiple vendors. Multivendor sourcing is now rearing its head in terms of complexity, and this is the primary reason why IT is so expensive.

PwC: You have led the development of virtualization solutions, and it is one of the key technologies in cloud computing developments. What role does virtualization have in addressing the concerns you have raised?

SC: The concept of virtualization is much richer than virtualization between an OS [operating system] and a hardware component, which is the most common use of it today. Virtualization is a layering function horizontally up the enterprise IT stack, between, say, servers and OSes, between OSes and applications, and between applications and end-user binding.

If you apply virtualization rigorously up the stack, you get an opportunity to cross those vertical silos at the appropriate level for your enterprise and examine whether it is your core competence or mission to manage the level below that particular virtualization abstraction. If it isn’t, then you can outsource it.

So, for example, if it is not your core mission to manage large data centers full of servers, then you can outsource that, focus on the things above it, and simply rely on the hypervisor as a way to separate and allow the IT stack to be dynamically bound with any set of servers run by somebody else.

PwC: It is common to see the cloud universe segmented into infrastructure-as-a-service [IaaS], platform-as-a-service [PaaS], and software-as-a-service [SaaS] layers. Which level will be more important going forward?

SC: Yes, there can be many layers. At the end of the day, I’m passionate about one layer, which is the bare-bones IT infrastructure-as-a-service layer. It’s extremely powerful, because it addresses some key needs for scale and cost efficiency, while not locking in the customer. That is, the notion of compatibility and portability is much easier to achieve than if you write an app against somebody’s platform as a service. It’s very important to preserve the notion of compatibility, and compatibility at the virtual hardware level is much easier than compatibility at the platform level.

“The enemy of agility is the existing vertical integration of expertise and function in the organization structures of IT.”

The other reason I like IaaS is to deliver multitenancy. In general, the application needs to be rewritten to get multitenancy to it in the cloud. If an application is offered as a SaaS solution and multitenancy is a feature, then you need to preserve that security and other abstractions all the way down through the application stack. It’s very, very difficult. On the other hand, it’s easy to provide multitenancy down at the hypervisor, network infrastructure, and virtual switching infrastructure level. Then you can have absolutely guaranteed separation from a security perspective, as well as auditability and compliance—without actually doing anything to your apps. You’d get there in an evolutionary way, and quicker, so I like that part.

PwC: Are we still very much in the early days of managing APIs [application programming interfaces] for a cloud virtualization infrastructure?

SC: In the [public] cloud, absolutely we are. In terms of the enterprise, VMware has the enterprise lead in virtualization, and the company has done an extremely good job of turning its administrator into the new hotshot of the data center and delivering a direct benefit to the business. In general, before virtualization, you had a bunch of people managing storage, a bunch of people managing servers, and so on.

With virtualization, the VM becomes a file. The storage person is just a provider of blocks who doesn’t do backup or DR [disaster recovery] anymore. It all happens out of virtualization. And the VM administrator is also the person who’s delivering the hard savings in terms of consolidation, power savings, et cetera. So this person is the new power elite in the data center.

PwC: As virtualization becomes pervasive, do enterprises create value and differentiation through the management of the virtualized infrastructure?

SC: Yes, though in general, I’m not wild about the term “management.” In the context of cloud and agility, I think there’s another abstraction that needs to be part of the IT infrastructure system. Once you have virtualization deployed, you need an orchestration layer that allows you to automate the deployment and running of applications. That is a key component of how you build this internal cloud function. I don’t think it’s management in the traditional sense of FCAPS [fault, configuration, accounting, performance, and security] style or getting alerts. That’s still drag and drop with the GUI [graphical user interface]. At the largest end of the scale, it’s all about automation, because that’s the only way that you get rid of the vertical sets of skills you must traverse every time you want to deploy something.

PwC: What is the current state of automation in the IT infrastructure in the data center? What capabilities lie ahead?

SC: The basic objective of automation is to get to the point where you have an administrative interface that IT can use to do its usual FCAPS-style management. You then expose to lines of business or users, according to their roles, a completely lights-out automated IT infrastructure where they can drive their own applications through their life cycle without requiring anybody in the IT infrastructure to flip switches, install software, or configure anything. That’s entirely automated beneath that line.

Where it gets challenging is in two areas: storage and network. The workload used to be statically configurable to the server, but now, with virtualization, the workload can move anywhere and that breaks the networking abstraction. Because the state [of the workload] is required to protect and enforce policies, the state must be agile too. Moving the workload also breaks the storage abstraction, because the unit of storage, which is the bootable entity of the VM, must be visible to the server that the workload is about to run on. Plus, you need to be able to perform backup, DR, and other activities there. Those layers of orchestration are required.

PwC: What will drive the adoption of cloud computing—greenfield solutions or the migration of existing solutions in the enterprise?

SC: Right now, four categories are relevant. One is surely greenfield solutions. Second is the stuff that you can throw up in the cloud very easily and not care if you lose it. Web servers and HPC [high-performance computing] scale up very nicely. There’s no issue around latency. You just want to consume some computing—you want to have easily scalable computing resources, and if somebody compromises a blob of binary data, they wouldn’t be able to use it anyway.

There are two other interesting categories. One is essentially DR as a category, so you have a site that can be used to recover if a disaster occurs. The other category is when an application is running locally, but if things go pear shaped [fails], it will be reconstituted in the cloud. Then there’s bursting to clouds. Now, you see bursting with HPC, but it’s happening already with our portfolio—XenApp and primarily cases that have some predictable workload and that have backup or additional entities up in the cloud, not running. When a flash of demand occurs, a portion of the traffic automatically redirects up into the cloud and fires up additional service for the load.

e see that one as a fairly important category. Arguably, it’s an extension of the same core technologies as the DR scenario, which has one instance of a typical workload, but not double the hardware costs, because you can have mission capacity on demand in the cloud. It’s just not running, and you don’t pay for it when you don’t use it.

PwC: All waves of technology adoption have been constrained by the legacy base of applications. What opportunity exists for enterprises to use cloud computing for their legacy workloads? Is there an opportunity to modernize the legacy base?

SC: The rate at which we lag—the tail is so long. When you go to the airport and check into any major airline, every piece of software that you interact with—from the time you walk in the airport until you get on the plane—was written in Windows 3.1, and that was written when I was in high school. That stuff is not moving. It can’t move; it’s too difficult. So there’s going to be this preservation notion of technology, which will adapt to those infrastructures. I think that’s a very, very interesting area to play in.

As per modernizing the base, how high up the stack do you have to go? Arguably, that’s Citrix’s core business. You have things that are legacy apps—strategic, where you cannot deliver them to end users—so instead of figuring out how to rewrite the app or anything else, you basically send a picture of the app from a server to an end user. That’s XenApp. You could take advantage of scale and various other things by doing it in the data center—reduced costs, reduced complexity, reduced management, and so on. Centralization reduces the number of things that you have to manage.

PwC: Given all that is unfolding with cloud computing, what should CIOs and their organizations be doing right now?

SC: The biggest barrier right now is lack of knowledge, and so I think people should just get going and do it. There are all sorts of issues. There are issues around licensing, scale, management, and everything else. But what doesn’t help is a lack of understanding of those issues and a lack of familiarity. I tell everybody that I meet who has had some budget canceled for buying new servers or not much demand from business: “Just find some stuff to throw out there. Go and use it.”

I think there’s a real knowledge gap to be filled. And because every vendor wants to be relevant, every vendor stretches the definition of cloud to include themselves. Again, that creates an opportunity for the CIO to analyze and articulate very concrete, very clear, concise benefits in the context of existing infrastructure that value-added providers can bring to customers today.