How Amazon cashes in on its Cloud
I had a very interesting conversation with Jeff Barr, the senior web services evangelist at Amazon. I have known Jeff for almost 15 years. In those days Jeff was one of the architects at a company called Visix, an early graphical development environment that was ahead of its time. Visix’s software development environment was designed as an abstraction of the underlying infrastructure. Visix came into the market before the Internet infrastructure became the defacto standard. But for me, it set the vision for where we are today. Jeff started at Amazon in the summer of 2002 with the Visix and some Microsoft experience in his consciousness.
Amazon’s business model is different than a traditional software company that often spends 18-24 months convincing customers to adopt new hardware/software or services. Amazon is leveraging a different computing model based on providing customers will a set of predefined services that can be bought without making a long term commitment. In a sense, Amazon has had the luxury (or good sense) to roll out service after service and see what sticks. As Jeff sees it, “people’s brains light up. They can build their business and applications in a positive way without having to worry about bandwidth, power and cooling.” His perception is that customers don’t think about whether the cloud provided by Amazon will support their needs. Clearly, he is able to talk this way because Amazon has made the investment in a scalable architecture to support an infrastructure that is designed for massive scalability. The other issue is that having built this architecture for its own retail requirements, Amazon had the foresight to exploit the technology to create a new line of business — in essence, a compute cloud based on providing a set of tools and product offerings to the market. The message to the market is straight forward, use these services so that you can innovate quickly without having to build from scratch. In taking this approach, Amazon creates both a test bed that allows the company to collaborate on new functionality with partners. In addition, and perhaps most importantly, it allows customers to buy incremental capability so they can scale up and down when they need to. According to Jeff one of the benefits of the cloud is that it isn’t dominated by the needs of one customer. In other words, one customer may have a spike in demand while another has less need at that point in time. Over the years, Amazon is able to understand usage patterns that are predictable.
Amazon’s business model follow this approach. A customer creates an account with Amazon that in essence gives them a charge account with Amazon. Customers get access to all of the Web services APIs. Their usage is tracked and they are billed for what they use. The business model is quite straight forward. Amazon charges 15 cents per gigabyte per month — not a lot of money even when you scale. What is interesting to me is there is no contracts to negotiate — everyone understands the rules. I asked Jeff if customers ever ask to buy a “private cloud”. While Amazon has been asked by customers, Jeff felt that because of the amount of experience that Amazon has with its hosted services discourages customers from explaining, “This is a business we want to be in. We have a lot of experience in our organization. We build highly cost effective data centers and sophisticated monitoring and operations.” He contends that Amazon has the expertise based on its 13 years in the business is enough to keep customers from walking away from its cloud. If you do the math, it would be difficult to argue. For example, if a customer needed 500GB of storage for two years, the cost would be $1800. In addition, it avoids the requirements for managing that environment.
Jeff makes a good point. If a customer needs to scale from 10GB to 100 TB in a month it might be hard to pull off. “This is routine for us,” Jeff claims. From his vantage point, the cloud changes the relationship between the customer and their hardware vendors. In effect, customers are sharing hardware resources with lots of other customers. So, the question becomes, who is your partner? It is no longer the provider of the hardware or the operating system. You probably still have a relationship with your software provider.
So, Amazon’s view of the cloud is pretty straight forward — it is a way to get value out of virtualization. Jeff points out that if developers uses Amazon’s elastic cloud service, for example, they pay to access servers on an hourly basis. Amazon allocates server to that account, provides a copy of the operating system they need to get started. That process takes a few seconds.
Another dimension of Amazon’s business revolves around the companies that actually build applications that sit on its infrastructure. Amazon has built a bunch of its own applications that it offers as services. In addition, there are a number of application companies that are building applications on top of the Amazon platform. One that Jeff mentioned to me is called RightScale, an automated cloud computing management system intended to help customers of Amazon’s Elastic Compute Cloud with issues such as load balancing. In addition to this type of company there is a community of 370,000 developers. Because Amazon sets the barrier to use so low, it is easy for a developer to try a service without making a long term commitment.
The more I think about Amazon’s platform and business model, the more sense it makes to me. I believe that Amazon and others such as Google, Salesforce.com, and eBay are a peak into the future of the new generation computing. In a sense, this model breaks every rule that the traditional computing industry has been built on. This movement towards enterprise software as a service and utility computing is beginning to redefine hardware, software, management and services. I predict that this new business model is going to slowly but surely turn the industry upside down. It isn’t only that the business model is different. The underlying technology platform based on standards and a service oriented architecture is propelling this change. The only thing that will slow this transformation is fear of change. But what else is new.