An increasing number of large enterprises are adopting cloud technologies, specifically private cloud technologies. In the most recent survey conducted by TheInfoPro, 22% of enterprises are investing in cloud technology, and over two thirds are spending the majority of their resources on private cloud development, says Carl Brooks, analyst, infrastructure and cloud computing, Tier1 Research, a division of 451 Research. As organizations retool the IT environment to create a private cloud, a private cloud architecture begins to take shape. It is important, however, that organizations consider their requirements for a private cloud and move forward accordingly.
“There’s a natural roadmap that organizations follow,” says Thomas Bittman, vice president and distinguished analyst, Gartner. “They virtualize to consolidate. Then they say, ‘now I have this pool of resources and I can be more flexible,’ so then they automate.” Automation allows organizations to provision virtual machines faster than a physical server. The third stage, says Bittman, is to build a private cloud. That involves creating a self-service interface on top of the automation.
But the organizations themselves don’t always distinguish between these stages. Some organizations say they’re running a private cloud when in reality they’re missing key components of the architecture. Bill Corrington, cloud strategy lead, Stony Point Enterprises, says he wonders when his clients say they are building a private cloud if that isn’t just talk for building another data center. “Is it really going to deliver highly elastic, scalable, pay for use services?” he says. “If you’re serious on delivering those capabilities and services to an internal customer base, then you need to have a management capability to handle usage tracking, billing, access control, making sure that customers can’t see each others’ data,” says Corrington.
Jason Bloomberg, president, ZAPThink, a Dovèl Technologies Company, concurs. “When building a private cloud, on the one hand you’re building a data center. You have to think about facility, power and cooling, racks, servers, the network, and you have to think about virtualization software. But even then you still don’t have a cloud. You have a virtualized hosting environment,” says Bloomberg.
Bloomberg encourages organizations to consider whether a virtualized hosting environment is sufficiently meeting business and technology needs, or whether it’s time to take the next step to the cloud.
“What is it that the cloud has that a traditional virtualized hosting environment does not have? The answer is elasticity. Dynamic provisioning and deployment of resources, and automated provisioning and deployment of services,” says Bloomberg. “On the private cloud side, it’s up to the enterprise to build all this for themselves.”
Considerations when architecting a private cloud
“The first [consideration] is obviously what exactly are they trying to achieve and what is the overall strategy for infrastructure within the IT organization,” says Brooks. This varies by the size of the company. It may be as simple as picking a goal and choosing the closest and cheapest service provider. When planning a private cloud architecture, consider whether it is for a single purpose and single application or part of a larger strategy for IT in general, says Brooks. Brooks also advises IT organizations to consider how big the private cloud needs to be and what else it might be used for. What tech stack will be used and how much do you want to pay for it, or will you build it yourself?
Bloomberg encourages organizations to consider whether they are looking for an operational production environment, or a development and test environment. The answer will impact scalability and bandwidth, he says. You should also think about elasticity. How dynamic does the environment need to be? A private cloud is limited in its elasticity because it is built from a traditional data center. “It is only as elastic as the size of the data center,” says Bloomberg.
Experts agree that the majority of private cloud architectures are being built to provide Infrastructure as a Service (IaaS). Bittman says these private clouds are based on virtual machines, and the virtualization technology that organizations choose often dictates which technology they use to take that leap from virtualized data center to private cloud.
When choosing a hypervisor to virtualize and consolidate servers, explains Bittman, that choice can be made in a vacuum. But that choice later influences what management and automation tools you choose. If you started out with VMware, you may very well choose to continue building a private cloud with VMware. However, there open source software is also a viable option.
Choosing cloud automation and management software
“The thing to think about when building a private cloud service, you’re basically building an interface for the user that completely abstracts everything around it,” says Bittman. He likens it to a restaurant menu. The customer sees what dishes are available, but they don’t know how it’s prepared. “Behind the scenes [of a private cloud], it’s all automated. If someone clicks item A, the provisioning, the deployment, the optimization, the ongoing management and updates – and the growing and shrinking – need to be handled automatically. You need automation down through the entire stack. This whole thing needs to be integrated and fluid from the top down,” says Bittman.
According to Bittman, Gartner breaks down the stack into five distinct pieces: access management, service management, service optimization, resource management and the resources. “We break it into these five pieces because vendors out there are moving into this private cloud market, and they have an offering and they declare victory. But they only have one piece. Most vendors have strengths in certain areas but weaknesses in others,” says Bittman.
“Today the primary concern is the immaturity of what’s available, whether it’s open source or commercial; it’s still very new. When we say we’re going to automate something, what we really want is to be able to push a button and it all happens automatically. What you don’t want is it to be almost automated,” says Bloomberg. “You don’t want to push a button, then have tweak a bunch of things,” he explains. “That’s what operations has done their whole careers.”
“In a cloud you don’t do anything. Now can you create that configuration sufficiently detailed so that you can do everything automatically? You deprovision and reprovision if there’s a problem with the configuration. Once you monkey with it, there’s no way to manage it anymore,” says Bloomberg.
Bloomberg offers this analogy: “In the past, we treated servers like pets. We gave them names and nursed them back to health. In the cloud, we treat servers like sheep. We number them instead of naming them, and if one gets sick we shoot it. There’s always one to take its place,” he says.
Ditch trial and error for frameworks
Slow and steady wins the private cloud race