The Why And How Of Private Clouds 2

IT pros may be getting sick of cloud computing terminology, but make no mistake: The private cloud is a new and powerful data center strategy. What IT leaders are sorting out today is just how much of the concept they want to embrace.

Consider Indiana University, which has about 1,300 virtual servers and select dynamic storage capabilities in what Dennis Cromwell, head of enterprise infrastructure, describes as a "fairly significant deployment of internal cloud services" at two data centers, at its Bloomington and Indianapolis campuses.

Yet the IT team decided to stop short of a self-provisioning model, like what Amazon.com and other public cloud services provide. For university staff, who request computing capacity via an online form, it looks much the same. But IT staffers then take several manual steps to provision capacity, usually within a half day. Cromwell says the university doesn't need a quicker turnaround, and there's an added governance benefit. "We continue to like to have a little bit of control," he says.

More than half of the 504 business technology professionals we surveyed say they're either using private clouds (28%) or planning to do so (30%). However, cloud computing isn't a single, shiny toy a company buys. Instead, it's a new approach to delivering IT services. It requires certain key technologies, but more broadly it focuses on standards and process.

Two critical characteristics differentiate private clouds from conventional data centers: the ability to pay only as you use services, and the ability to scale usage up or down on demand. Tactically, it includes pooling of computing and storage resources to provide that elasticity. Some people also consider self-provisioning, where end users can allocate their own computing capacity, part of a private cloud.

Cutting IT costs is the No. 1 driver, but nearly as important is the ability to quickly meet user demand, our survey finds.

Here's a dash of perspective on that 58% using or planning to use private clouds--few if any of those companies have bet the business on it. Only 15% devote more than 20% of their IT budgets to cloud computing. A third say they spend 10% or less, while 38% have no defined budget for it. Creating a private cloud takes a technology investment, but only 24% of those with no plans for private clouds cite excessive costs as a barrier to cloud computing.

In fact, there's no dominant obstacle--the most-cited barrier (37%) is "no business need." Security concerns come next, followed by legacy systems and processes that don't fit the model. This mix of barriers, rather than any one big problem, fits with our conclusion that most IT organizations are experimenting with private cloud technology and figuring out whether and how to apply it.

A key concern is whether private clouds will provide long-term savings, particularly whether that up-front investment in data center efficiency will deliver sufficient economies of scale for a single business, even a very big one, when compared to the massive scale in a public cloud environment.

Business technology teams should expect to find plenty of fragmented products, and they'll likely struggle to put all of the pieces together to create a true elastic, on-demand private cloud. However, many of the steps you should be taking to run a better data center--like increased automation and consolidation--will help lay the groundwork for private clouds.

That's what the U.S. Army is doing: pushing two major data center initiatives around private clouds and data center consolidation in lock-step (see story, p. 28). Indiana University is doing a complete server refresh as part of its private cloud effort, tapping into the higher performance and efficiency the latest boxes offer. The state of Utah is using a private cloud project to ramp up its server consolidation. These moves make sense, as optimizing and consolidating data centers are needed for a successful private cloud.

Automation and service management software are also critical, as is a high level of server virtualization and standardized processes for tasks such as provisioning capacity. Large organizations in particular are adopting elements of cloud computing such as Web-based service requests, dense virtualization, and data center automation. All good steps. Just be sure to have a complete strategy before getting into the private cloud in a piecemeal way.

Private Cloud Momentum

Here are five attributes, outlined by the U.S. National Institute of Standards and Technology, of cloud computing services: On-demand self-service, ubiquitous network access, pay for use, elastic scale, and resource pooling.

But don't get bunched up over definitions. The bottom line is that companies see opportunity in the flexible, pay-as-you-go capacity of public clouds offered by the likes of Amazon, Google, and Microsoft, but they're nervous about losing control with those services. Spurred by that desire for data control, governments are emerging as leaders in implementing private cloud projects. The Department of Defense Information Systems Agency, for example, has developed on-demand server space that development teams can tap in 50-GB increments, with prebuilt security standards.

So while neither public or private are widely adopted today, companies appear more likely to put money into private clouds in the coming years, our research finds.

In our survey, 63% of respondents have no plans to use public cloud services, while 20% use them. Looking ahead, only 17% have plans for public clouds, compared with 30% for private clouds. Spending for private clouds, while low today, looks poised to ramp up: 18 months from now, one-fourth of companies in our survey expect to be spending more than 20% of their total IT budgets on private clouds.

While we noted the barriers to private clouds, it's interesting to learn what IT teams aren't worried about. Only 8% worry that the technology's inadequate, and just 9% fret about IT organization resistance--though they're worried about having the horses to carry it out (24% cite lack of manpower or skills).

IT teams may be underestimating some of the risks. The technology's not as mature as the survey lets on, and private clouds are by no means a cheap move that'll sail through budget approvals. Since cost savings come from scale, getting as much of the company as possible to buy in is vital to determining private cloud success.

Private clouds don't make sense for small businesses. But at large and even midsize companies, IT teams should make private clouds a rallying cry to garner support of efficiency projects.

The Right Technology Changes

Private clouds won't be built on niche products: 55% of companies will look mostly to their existing vendors. Yet there's room for some newcomers; 41% don't yet know which vendors they will use.

IT organizations building their own clouds need to make sure core data center technology--including servers, storage, security, virtualization, and network infrastructure--are capable of the performance needed. However, there's not much new there compared with conventional data center technology, except perhaps virtualization. And even that's extensively implemented in most shops. The major area for investment, since it's lacking in many environments, is true service management and automation.

That's our take, but IT most pros don't see it that way. Nearly half of survey respondents expect their servers will need updating for private clouds, 45% think storage will, but only 20% think their systems management will, and just 14% their provisioning software.

It's true that companies will want quality hardware performance, and it's possible that will justify upgrades. Indiana University, for example, decided that its move to a private cloud architecture justified a complete server refresh. That was based on the performance and energy gains, very aggressive prices from vendors, some cost savings related to licensing, and the advanced age of many of its servers--though it also replaced some servers just a year or two old. "You do need to standardize on a small set of physical servers," says Cromwell. The giants of public cloud computing custom-design servers to eke out every bit of performance, eliminating unneeded peripherals that may add a few dollars or watts. Increasingly, server vendors are offering such cloud-friendly servers for the mass market.

Companies may need to upgrade their networks to meet the demands of larger file systems moving throughout the data center, or to meet redundancy requirements. For enterprises, on-demand scale is the toughest problem to solve, and you still need to perform capacity analysis to build in the appropriate amount of surge resources.

Storage may also require some investment, to allow for more flexible pooling of resources. NASA, for example, is considering using its Nebula cloud environment to dynamically provide storage and memory for the virtual machines of one of its mission groups.

The state of Utah shows how a private cloud project can spur consolidation efforts. Even after four years of data consolidation efforts, the state's applications resided on 1,800 physical servers in several dozen locations. Now it's moving to a private cloud and is nearly 75% through a virtualization migration, that together will reduce the server count to 400 with savings projected at $4 million annually.

It's on the software end where companies will need to pay the most attention, and money--37% of companies expect to have to spend more on virtualization to make their private clouds a reality.

Inventory management software is also at the heart of the private cloud software stack, monitoring all aspects of private cloud resources: the operating systems on each machine, physical servers, space, storage, and other infrastructure. On top of the inventory system is usage and capacity management. Finally, there's the orchestration software containing trigger points when more resources need to be added.

Six Steps To A Private Cloud

While technology choices are important, a lot more process, policy, and financial analysis work needs to be done beforehand. Here are six steps to develop a private cloud strategy.

1. A service catalog. If you aren't already, you need to think about your IT infrastructure as services. There's software to help create a service catalog, and that will be vital at some point to running a private cloud. But you can get started with a list describing services and the IT components behind them, along with supporting business information, such as the performance needed and the cost. The catalog lets IT manage demand, profitability, and performance. But it's not a small task if a large company's still thinking about systems as individual components.

2. The business case. Since many companies will convert existing data center services to a private cloud, CIOs need a handle on today's total costs plus the planned growth over the next two years.

One complication: It doesn't make sense to do a neat, drawn-out, service-by-service migration to a public cloud. You can't deliver scalable capacity that way, and you won't get the savings from economies of scale. Creating a private cloud for a small group of users or services will likely be needed as a test case, but you should plan to scale up quickly.

3. Standard operating procedures. IT needs to clearly define who has access to systems and the policies and procedures for provisioning and deprovisioning systems, resources, and access. While manual provisioning works in some environments, to get full benefits of the cloud environment, IT teams should push for automation to reduce errors, especially for changes in configurations.

From an operational perspective, standard processes such as ITIL are needed in private clouds to drive the efficiency and allow the automation needed to achieve the cost savings. Plus, when problems occur, the rapid triage of these standard processes is a must for maintaining credibility. Key process areas to define include incident, change, release, problem, and configuration.

4. Standardization and shared use. Public clouds have scale. Microsoft, in the massive data centers it uses for its Azure cloud service and software-as-a-service applications, runs more than 3,000 servers per data center operator.

For a single company, though, the pool of users is much smaller, so you need to ensure that every part of your company will use the private cloud service. If you have holdouts, that needs to be resolved before moving to a cloud environment. A shared-use policy, backed at the executive level, is a must. Shared use requires standardization not just of IT operations but also around elements such as acceptable service levels, application types, and capabilities provided by cloud services. Private clouds can even act as a catalyst to consolidate applications, hardware, and infrastructure in the data center.

5. Interoperability. Surprisingly, few people are concerned about compatibility--only 18% in our survey think compatibility with public clouds is important. So-called hybrid clouds hold little interest, with 69% having no plans for them. Yet even for a company that doesn't plan to tap a public cloud, what if you acquire a company that uses public clouds or its own private cloud?

Cloud interoperability is very difficult today. Many public clouds use Xen for virtualization, and most enterprises run on VMware. We're seeing some emerging technologies to enable federated cloud environments and broader API support, but the platform support is narrow. Organizations such as the Open Grid Forum Open Cloud Computing Interface are working on an API for remote management of cloud computing infrastructure, which would let tools work across clouds for common tasks such as deployment, scaling, and monitoring. Today, each cloud has its own APIs.

Until the industry sorts out which will be the long-term fix, look to places such as the Cloud-Standards.org wiki, dedicated to the dozen or more groups working on interoperability.

6. Technology. Here, your overall sourcing model comes into play. Do you want to build and run your private cloud, or outsource it? IT service providers are jumping into this market. For example, Hewlett-Packard last week said it's spending $1 billion over the next several years to automate its customer-facing data centers, in part to prepare them to deliver cloud services.

At each of these steps, you'll need to look at security and compliance. Asked the importance of different private cloud features, "highly secure" is cited as very important by 65% of the IT pros in our survey. That's probably not much different from what IT pros would say about their existing data centers, but a private cloud does create new compliance concerns. Since workloads move around from physical servers, even if they're inside your data center, you may have to invest in new reporting software to meet auditing rules.