The hidden costs of a private cloud

The phrases "private cloud" and "public cloud" have become commonplace, but when you think about the benefits of a real cloud computing environment, the hype around a private cloud doesn't measure up

The term "cloud" has attached itself to practically every product or service offering compute and storage. This enormous popularity of the cloud is not driven by the coolness of technology but by the clear and obvious benefits that cloud computing gives its users, which is revolutionising the way IT services are being procured and used. The phrases "private cloud" and "public cloud" have become commonplace, but when you think about the benefits of a real cloud computing environment, the hype around a private cloud doesn't measure up, writes Werner Vogels, chief technology officer at Amazon.

Cloud computing means that instead of buying, owning, and maintaining datacentres or servers, a business or developer purchases IT services, such as compute power and on an as-needed basis. Database, messaging infrastructure and content distribution services are also available in the cloud. The provider manages and maintains the entire infrastructure in a secure environment and users interact with resources via the internet. Capacity can grow or shrink instantly.

For a service to truly be classified as cloud computing, it must require no up-front capital investment. Services are pay-as-you-go and provide a flexible pricing model. Developers have instant access to server capacity and can quickly bring projects to market. Capacity is truly elastic, so businesses can scale up or down any time and they never sit on or pay for unneeded, excess capacity. Another important feature of cloud computing is how it allows businesses to offload the undifferentiated heavy lifting of infrastructure so engineers can focus on projects that define a business. And finally, the infrastructure in the cloud must comply with regulatory needs without sacrificing data security.

If you look at how businesses deploy what many call a private cloud, the benefits listed above are lacking - especially in the financial category. In a private model, businesses still own datacentre space and servers. That requires a fixed, up-front investment. Capacity is limited by the amount allocated by the owned infrastructure, so it is not truly elastic. Services are not pay-as-you-go because you have already consumed large up-front costs for the hardware. Running large scale, high availability infrastructure requires highly innovative and motivated staff and the attention of upper level management. This represents a lost opportunity for enterprises to focus and innovate in their core businesses.

Many organisations are surprised by the costs involved in managing datacentre infrastructure. To truly understand the cost-burden of IT infrastructure, people need to understand the following factors:

Utilisation: The average datacentre utilisation is around 10%, and 20-25% are the highest numbers we hear. An experienced cloud provider will run at a much higher utilisation, as the demand curve is smoothed across many thousands of customers. This means that customers don't have to pay for idle time, which they are doing in their own datacentres because they own their servers 100% of the time and have them turned on 100% of the time.

Redundancy: Running a highly reliable and available IT infrastructure of course requires far more than deploying servers in a datacentre. Organisations need to have reliable storage, including backup. This goes beyond RAID, given the average failure rates inherent in any single storage array or device. They must also build and operate a reliable network, by provisioning redundant networking devices, transit connections and physical connections between datacentres. Beyond backup and reliable networking, they must have a tested, working solution for disaster recovery. This includes deploying in multiple datacentres - either with failure-resilient software or in a more traditional hot/cold standby approach. To achieve realistic disaster recovery, all of the datacentres and servers involved have to be constantly utilised; if they sit idle, it is almost certain they won't function as desired when activated from a cold start. So an organisation needs to account for both the cost and the complexity of this redundancy when evaluating deployment.

Hardware costs: Beyond the initial investment in purchasing hardware, businesses need to account for performance checks and maintenance fees. Upgrading to the newest routers or load balancers is expensive. For customers requiring a multi-million-dollar investment in infrastructure, thinking about how the value of that asset will degrade over time is a critical analytical component.

Supply chain: It is extremely common to incur very significant over-capacity caused by the time that passes from when hardware is ordered to when it is brought online - often running many months. With such long approval and delivery cycles, excess capacity spreads throughout the pipeline and all this unmeasured waste is not captured. Dedicated cloud service providers spend significant resources managing all these aspects of the supply chain well, in order to drive down costs for the customer.

• Datacentre efficiency: The Uptime Institute reports that most datacentres have a power usage effectiveness (PUE) of 2.0. This means that for every watt that is delivered to the servers, one watt is wasted in overhead. Heavily invested datacentres are far more efficient. You need to have a lot of servers to afford to devote the resources needed to run the supply chains at maximum efficiency.

Personnel: Many people don't think about the true costs of the sizable IT infrastructure teams that are needed to handle the "heavy lifting" of managing heterogeneous hardware and the related supply chain, staying up-to-date on datacentre design, negotiating contracts, dealing with legacy software, operating datacentres, moving facilities, scaling and managing physical growth. To get specific:

  • Hardware procurement teams are needed, who have to spend a lot of time evaluating hardware, negotiating, holding hardware vendor meetings, managing delivery and installation, etc. It is expensive to have folks on staff with sufficient knowledge to do this well.
  • Datacentre design and build teams are needed to create and maintain reliable and cost-effective facilities.
  • Operations staff is needed 24/7/365 in each facility.
  • Networking teams are needed for running a highly available network. Expertise is needed to design, debug, scale and operate the network and deal with the external relationships necessary to have cost-effective internet transit.
  • Security personnel are needed at all phases of the design, build, and operations process.

The cloud is not simply a way of saving money. It gives businesses opportunities to be more productive, agile and responsive than is possible with physical hardware in a customer's own datacentre.

We make these distinctions because it is important for people to see the big picture and understand the business impact and realities of a real cloud environment when making decisions for long-term IT infrastructure plans.

For many enterprises, integrating cloud computing into their IT operation is a multi-year process that requires careful, strategic planning. We have found cloud computing is a no-brainer for several classes of applications including web and application hosting, collaborative tools, online marketing campaigns and data analysis. Other classes of applications require more planning.

Most companies have a number of existing applications that were not written in a way that can easily take advantage of the cloud and its elasticity. Additionally, some applications have dependencies on other internal applications or data sources that need to move to the cloud too before they can function correctly At AWS, we recommend building a migration plan, usually spanning 24 months, and transitioning application by application. This enables organisations to gain experience with the cloud as they begin to transition larger chunks of their infrastructure.

CIOs have tough jobs and are responsible for the performance, security and reliability of thousands of applications. Relinquishing control is understandably difficult and there is a comfort in the ability to immediately walk down the hall to a server room and take action if there is a problem. The truth is a cloud computing environment such as AWS can give IT managers a great deal of control and visibility, allow them to choose where data is stored and provide the flexibility to back up their applications with the same security measures as before and with the highest levels of reliability and performance.

As more and more enterprises transition to the cloud, we expect to see heightened innovation as companies break through old world IT constraints.

Read more on Data centre hardware