CIOIN

Few IT crazes have reached the fever pitch that cloud computing has attained. Almost from day one, intense excitement has greeted the vast potential offered by enormous, hyperscalable public clouds that can scale up and down while customers pay only for the resources consumed.

Despite this excitement, however, just a small slice of enterprises are actively using the public cloud for core, mission-critical functionality. The reason? Mainly, widespread trepidation about the level of security, reliability, and data portability that current public cloud offerings can provide.

The industry’s answer to these concerns is the private cloud. The private cloud attempts to offer the same self-service agility and scalability that public clouds offer, but without the complications of putting critical services and data in the hands of a third party. But as with any sweeping IT concept, many customers are having a difficult time grasping what a private cloud really is and how they can benefit from building one.

Yet a common thread of shared resources, more efficient management, and greater business agility unifies all private cloud implementations and can be applied to any size environment.  

Two Views of the Cloud

As always seems to be the case in IT, at least two different perspectives surround the adoption of any new datacenter technology: The view from the CIO’s desk, which largely concerns itself with business goals; and the view from within the datacenter, which is driven by technology and the struggle to manage an ever-expanding workload. The good news is that the private cloud, when implemented for the right reasons, can meet both sets of needs.

A CIO might see the private cloud as a means to deliver better service levels, improve responsiveness, and allocate resources among business units more effectively.

Also, the newfound agility and efficiency of the private cloud can decrease the likelihood internal business units will “get tired of waiting for IT” and adopt public cloud services willy nilly–weakening the IT organization, creating new silos and redundancies, and opening potential security vulnerabilities.

A 2010 survey of IT decision makers conducted by Forrester Research concluded that only 13 percent of enterprises surveyed were using cloud-based IaaS offerings, but Forrester believes the true number to be nearly double that. “It often comes as a big shock to the infrastructure and operations people [within IT] to find they grossly underestimated the cloud services in use at their organizations,” says Galen Schreck, Forrester vice president and principal analyst, “They realize they have no idea what the application owners [in business units] and developers are up to.”

That’s a dramatic statement and CIOs are definitely taking notice. Day by day, they risk losing control of their organization’s data–data they are ultimately held responsible for managing and protecting. In the old days, “rogue” projects typically took the form of departmental servers hiding underneath someone’s desk; today, data migrates to third-party public cloud providers without planning or oversight, risking data loss or regulatory violation. To the CIO, the private cloud seems like the silver bullet to stop this.

But a full-scale private cloud doesn’t just require technology, funding, and know-how. It also requires a number of changes in the way IT is run on a day-to-day basis. Attempting to implement a private cloud without business acceptance of a chargeback funding scheme or resource pooling is just as counterproductive as providing business units with the technology to provision their own server resources in an environment where they may not have the skills to take advantage of it.

What Makes a Private Cloud?

Before delving into what a private cloud looks like from a rack and sheet metal perspective, it’s important to understand what problems a private cloud is designed to solve and how that sets it apart from a traditional, even fully virtualized, on-premise  infrastructure.

Agility

Business units like to complain to CIOs and IT practitioners that it always takes too long to provision new services, and they often decry the up-front cost associated with them. A business unit seeking to deploy a new application may spend months or even years deciding which software vendor to purchase from and lining up development resources and consultants. But once contracts are signed and plans are put in motion, business stakeholders expect IT to react quickly and fulfill infrastructure needs.

For a traditional IT department, unanticipated requirements can be extremely difficult to manage. Business stakeholders often underestimate the server, storage, and data protection resources that their new application will require, and they may not account for the time it takes to order, receive, configure, and implement. Alternatively, the contract for the software may have included hardware intended to be dedicated to the new application. In the latter case, not only will IT be saddled with managing that hardware, there’s also an excellent chance the software vendor will have massively over-spec’d it–resulting in even less operational efficiency.

At best, this process is an expensive waste of time. At worst, it can have a lasting negative impact on the working relationship between the business unit and IT. It’s easy to say that the solution lies in better communication between IT and the business. That helps, but very few IT organizations manage to fully cross that chasm.

The private cloud essentially allows everyone to have their cake and eat it, too. Project sponsors can access various types of server and storage resources that IT has made available through a self-service portal. They can review the specifications and costs of each and share them with the software vendor, which can make recommendations on which they should choose. When it’s time for the application to go live, the business unit “orders” the services, which are automatically provisioned and immediately available for use, all without IT needing to do anything or even necessarily be involved.

Configuring the portal, policy, and automation magic that makes it all work requires time and effort. But the efficiency benefit can be big, especially when system provisioning is a common task.  

From a political standpoint, the benefits are much more obvious: IT is no longer a speed bump for the business units but still retains control over the infrastructure.

Scalability

Another key requirement placed on any IT infrastructure is the ability to quickly scale in the face of increasing load. Traditional IT generally handles this by over-provisioning infrastructural resources as they are purchased by business units. This gives IT some cushion before stakeholders will demand additional resources to cope with higher load.

Yet over-provisioning contributes to the perception among business stakeholders that IT is too expensive. Plus, this approach fails to scale beyond a certain point, after which yet another round of costly capital expenditures ensues. Worse, by repeatedly over-provisioning small islands of dedicated infrastructure, IT strands large amounts of capacity and prevents those resources from being used to satisfy spikes in demand elsewhere.

When these application loads live inside a private cloud and business units are paying on a per-usage basis, IT no longer has to dedicate resources to each business unit individually. Instead, they can pool the entire corporate infrastructure–servers and storage–and manage a single pool of spare capacity. It’s easy to see how this can decrease overall costs. Just as business units can deploy a new application with little lead time, they can also increase the amount of resources granted to one that they have already deployed–even to satisfy a short-term increase in load–and then contract them afterward.

Multi-Tenancy

One of the few good things about traditionally deployed dedicated infrastructure is that it’s fairly easy to maintain divisions between the infrastructure serving various applications and business units. These divisions may simply consist of installing applications on different servers, providing security and performance segregation. But they may also extend all the way down through the network and storage infrastructure. Such physical separation allows IT to implement a high degree of security easily, but it also results in an incredible amount of waste.

Although resources are pooled on the same server, network, and storage hardware in a private cloud, IT must still maintain appropriate performance and security segregation between the various workloads for the resulting product to be acceptable to business units. This segregation is accomplished through automatic configuration of the virtualization, network, and storage hardware as the services are provisioned.

During the provisioning process, the automation engine will build out a virtual machine with processor and memory allocations, limits, and reservations that match the specifications the business unit chose for the system. In addition, it will automatically configure a secure network for the system, generally using a software-based firewall for edge security. It will also, based on policy, configure the storage for that virtual machine. 

Although the level of direct storage integration varies from product to product, the service level for storage can be based on either known service levels for various pools of storage that users can choose between, or, ideally, on actual service-level configuration within the back-end storage itself.

Governance

In traditional IT environments, IT governance–really just an explicit set of policies–is often seen as an obstruction in the path of business units seeking quick deployment. The lumbering nature of governance often derives from the fact that IT must apply and reapply the same policies over and over as each business unit brings in a new application or upgrades an old one.

How much performance is required? How will the data be stored? What are the data retention policies if a system is decommissioned? What kind of redundancy will exist? These are only a few questions that IT really must ask to do its job, but the business unit almost always sees them as obstructionist and, above all, expensive.

The private cloud does not eliminate these governance requirements, but it does simplify them by allowing IT to effectively answer them once for the entire shared infrastructure and build those infrastructural costs into the usage fees that business units pay. 

Remember that provisioning within a private cloud is driven entirely by the business units, largely without direct IT involvement. Therefore IT must be very careful when it constructs the policies that define the different compute and storage products business units can choose from–and in defining the SLAs attached to them. So although IT can improve its own customer relations by requiring less information from business units, it also has a far larger internal policy burden to bear.

Storage

Storage is the bedrock of any IT infrastructure. At first glance, it seems that deploying storage for a private cloud would be relatively simple. Instead of requiring a fleet of different storage resources, each dedicated to a different business unit or app, a single integrated storage pool can shoulder the load of the entire cloud infrastructure.

True enough, but that fact alone demands storage solutions specifically tuned to accommodate a large number of disparate workloads. The storage must be able to scale extremely easily, must be capacity-efficient, must manage performance and tiering autonomously, and, ideally, should be easy to integrate with cloud management software. This is a tall order. Only a few storage products satisfy all or most of these needs.

Regardless of what kind of storage is used, monitoring and managing storage capacity and performance levels are extremely important in private cloud environments. Since IT may not have any warning that large influxes of new workloads are going to spin up, it needs to be able to turn on a dime to add additional capacity.

 Failure to adequately manage storage capacity and performance in a private cloud environment can have far-reaching impact on a wide range of users. That storage, after all, is a huge pooled resource.