In part one of this look at private cloud, we examined what it takes to build an in-house cloud service. In part two, we'll find out why all enterprises aren't hopping onboard the private cloud bandwagon.
Given the hype and, possibly, executive pressure to implement a private cloud infrastructure, many companies have struggled to justify the massive organizational change and expense. While many IT managers believe that their shops aren't yet ready, their objections may fall on deaf ears among executives who care only about the bottom line. Still, if you are an IT decision maker, skepticism about cloud computing is not only warranted; it's essential. The size of your company and other key factors -- such as the degree of virtualization in your IT shop -- should dictate whether your company is ready for the move.
If done properly, this new [cloud] model changes IT's seat at the table.
Mike Pearl, head of cloud computing at PwC
Large companies with hundreds of departments all placing heavy demands on IT are good candidates for a private cloud. The self-service, automation and chargeback capabilities of a private cloud make a huge difference, but only at scale. This is because the economic benefits of moving to a cloud model are achievable only when all or most IT services run on a cloud platform and where the use of compute resources is high. Simply put: The bigger the cloud, the more cost-effective it is.
In addition, large companies that are heavily regulated and fearful of moving data to a public cloud are also good candidates for a private cloud. Organizations that have made a significant investment in virtualization and in running large-scale virtual environments are well-equipped to move toward a private cloud model.
According to the Microsoft white paper "The Economics of the Cloud," for small and medium-sized organizations with fewer than 100 servers, private clouds are prohibitively expensive compared with public cloud services. The only way for these small organizations or departments to share the benefits of at-scale cloud computing is by moving to a public cloud model, the white paper argues. For large organizations with an installed base of approximately 1,000 servers, however, private clouds are feasible. But they still come at a significant cost of about 10 times the price of a public cloud for the same unit of service given the combined effect of scale, demand diversification and multitenancy.
And the questions of scale and cost are just the beginning; management software immaturity is also a major hurdle. Thomas Bittman, an analyst at Stamford, Conn.,-based Gartner Inc., notes that technology may be the easiest challenge to overcome in building private cloud services, but the market for private cloud technologies is still nascent. Numerous companies -- from traditional infrastructure and management vendors to virtualization and focused vendors to startups -- are developing products that focus on cloud management. This crowded field only complicates the vendor selection process.
The anatomy of a private cloud: Lessons learned
At some companies, the stars of cost, scale and the degree of virtualized infrastructure have aligned to justify a private cloud project. But even with a primed infrastructure and a sizable IT environment, companies struggle to get a private cloud off the ground.
Existing processes. After building a private cloud at his global financial institution, virtualization architect Stuart Radnidge admits that the process isn't for the faint-hearted. According to Radnidge, it mostly involves changes to your current operating model. "There are things that just pop out of the woodwork that are completely unrelated to cloud computing but are mandatory in order to move to a cloud environment," Radnidge said.
An important first step, he said, is to identify why you are moving to a cloud model in the first place. Is it to provision IT services faster, or is it to save money on server racks or power and cooling or staffing costs? Governance is another important and often complex issue. What is the cause of these problems? How many of these issues can be solved by new technology versus by introducing new processes?
Depending on your goals, the path you take to a private cloud may be radically different. It may be enough instead to reevaluate your existing processes, although this, too, can be a tough road. If the problem is cost, is it absolute cost or value for money that's the problem? You can often address a value concern by process improvement instead of by introducing new technology.
Tools of the trade. Radnidge cautioned IT managers against products that are mostly smoke and mirrors.
"Cloud solutions can give the appearance of intelligence, but if it's just a bunch guys running around in the back doing stuff in the same way with the same old manual processes, it will more than likely come back and bite you," he said.
Then there is the question of using existing tools and whether the usual suspects will suffice. Radnidge said reporting and monitoring tools are crucial to a private cloud implementation. It might be possible to bend existing tools to work in a cloud computing model, but these tools may just break. Still, he said, you may need only a few discrete components to enable your internal private cloud vision.
Unfortunately, few companies out there offer discrete components: "They all want you to buy into their complete stack," Radnidge said. This isn't necessarily a problem if there is a component you absolutely need and have no other way of delivering. He urged organizations starting down the road to a private cloud to develop an application programming interface (API) first rather than trying to tackle this later.
"Keep a close eye on what external cloud providers are doing, as it is inevitable that at some point you will be going external," he said. "It will make life a lot easier if your internal APIs are not too different from what the external APIs look like." He advised enterprises to check out APIs from VMware Inc., Rackspace Inc., Amazon Web Services and the major open source projects as well as the provider extensions from companies such as Terremark Worldwide Inc. and Savvis Inc.
Vendor-company partnerships. Given the immaturity of cloud computing, you can benefit from other companies' mistakes and learn from their laboratories. Radnidge advised IT departments to watch external cloud providers, which are moving faster than enterprise IT. He suggested forming strong partnerships with providers to give them your requirements, which will benefit you down the line when the day comes for your organization to make the leap to public cloud.
Cultural changes needed
From the Forrester Inc. survey, which surveyed 2,000 IT decision makers from North America and Europe, it's clear that there's a long way to go before most companies are ready to build a private cloud. One of the biggest barriers is cultural. While historically IT has largely enabled operational efficiency, today's IT workers need to focus more on bringing strategic gains. Some may resist the changes inherent in a cloud model in the name of organizational security and control. But this resistance often stems from an old-school view of IT resources as assets to be doled out by guardians of the "fortress."
Keep a close eye on what external cloud providers are doing; at some point you will be going external.
Stuart Radnidge, virtualization architect
According to Mike Pearl, head of cloud computing at PricewaterhouseCoopers, a cloud model requires a shift among IT managers from being custodians of technology assets to service orchestrators of technology that can run internally or externally. If IT departments don't come to view end users as customers of a service, these users will eventually turn to external services instead. While IT departments may fear that the shift toward the cloud model threatens their jobs, for the first time, in fact, this change should empower IT managers.
With the right skills to implement and manage a virtualized infrastructure, many IT managers may better ensure the longevity of their careers rather than risk obsolescence. David Foote at Foote Partners LLC predicts that while many "generic" data center jobs could disappear as companies make the shift to the cloud model, there is plenty of room for IT specialists in converged infrastructure systems, such as Cisco's UCS, as well as in networking and security, and more. Foote predicts that technologies such as virtualization and cloud computing will create only greater needs for these skills, not force IT underground.
"If done properly, this new model changes IT's seat at the table; it becomes a much more important seat," Pearl said. "It is a daunting challenge and there is an enormous pressure on CIOs and IT operations to get this right."
But getting it right, as Radnidge's experience indicates, is an iterative process, and cloud computing is a long-term strategy that may require several cycles to develop.
Jo Maitland is the Senior Executive Editor of SearchCloudComputing.com. Write to her at firstname.lastname@example.org.
This was first published in May 2011