For many, cloud computing and Amazon Elastic Compute Cloud have become nearly synonymous terms, and all other cloud providers are just trying to catch up. Is that really the truth? Is Amazon actually ahead or behind?
Most financial analysts place Amazon Elastic Compute Cloud, or EC2, at the top of the cloud provider list in terms of total revenue. Everyone else, according to analysts, trail behind. But Amazon is an online retailer -- a company that earns only a small chunk of its total revenue from the cloud.
So, why is Amazon getting all the attention? And will big cloud consumers really bet on cloud services from a retailer? A large part of this answer depends on the type of services potential consumers plan to put into the cloud.
IT spending is about one trillion dollars worldwide, about two-thirds of which consists of core business applications that support workers, maintain records for billing and accounting, stage materials for manufacturing, and more. The remainder of this spending is divided among a large and fuzzy group of activities that include email and Web hosting, business intelligence (BI), and personal productivity support.
In this core applications group, Amazon and other early market leaders such as RackSpace and Microsoft Azure have targeted test and development of critical projects. Creating a flexible Infrastructure as a Service platform on which companies can host pilot and test projects allows them to isolate test systems that might be hit by software bugs from critical applications. Amazon has a clear lead here, and one that might be difficult to erase, even for huge competitors with deep pockets.
Now, Amazon and competitors like RackSpace are targeting a second application group, the "hosting plus" space, as it's been deemed.
While most large enterprises tend to host their own websites because it's easier to integrate Web commerce and customer information with core applications, many smaller businesses pay for Web hosting services. These companies build a type of Web front-end to core applications; Amazon EC2 is a popular choice as this connecting technology. Many Web startups that rely on venture capital, for example, host their sites on Amazon to help minimize costs.
For prospective competitors to gain ground on (or spring ahead of) Amazon EC2, they may need to rethink their pricing structures. For example, Amazon EC2 charges for computing, storage and for traffic served onto the network. That's the same basic price model of many Internet hosting providers.
The prevailing cloud service pricing model charges for data storage as well as outbound data traffic at a rate of $100 or more per terabyte per month. This model may be suitable for pilot projects, Web hosting and email applications, all of which use modest amounts of data (a few hundred gigabytes). However, this pricing structure would actually double the cost of core applications, because not only does the average enterprise have a thousand or more terabytes of repository data, their applications read and write that data on a daily basis. After a month, the traffic from a repository store in the cloud could cost tens of thousands of dollars.
It is this core data pricing model that gives Amazon competitors an upper hand. Disk devices cost must less than $100 per terabyte to purchase, so a highly efficient data center that's designed for low-cost storage could host core application in the cloud at a reasonable cost. Large service providers, including cable operators and telcos, for example, typically see a low return on infrastructure investment, so they could price an efficient data center aggressively, perhaps so aggressively that Amazon and other large cloud providers might be reluctant to match.
Will big cloud consumers really bet on cloud services from a retailer?
Placing the right app in the cloud
Cloud providers can also use other application-specific strategies to make it more affordable to run core applications in the cloud. One strategy is to target applications like BI that aren't run directly on large data repositories, but reside instead on smaller summarized subsets or data views. Even though the whole database might be too large to host in the cloud, the abstracted summary is typically smaller and can be a better fit.
Another strategy is to allow cloud applications to access data that's still stored in enterprise data centers. This could be particularly effective in enterprises that turn to the public cloud for additional processing capacity during peak load periods, or cloud bursting. This approach could also prevent an enterprise from incurring cloud provider charges for network data use, since most of these charges are levied only on outbound data.
Both of these approaches would require that the cloud user design an application to optimize cloud data costs, which could increase the level of business and application support necessary to sell and set up the cloud service.
ABOUT THE AUTHOR
Tom Nolle is president of CIMI Corporation, a strategic consulting firm specializing in telecommunications and data communications since 1982.