Where Microsoft built Azure largely from scratch, Amazon already had the basic infrastructure in house to support its massive e-commerce operations. Always eager to expand, company founder Jeff Bezos and other execs saw an opportunity to commercialize the technology that was serving its business so well.
That was 2003.Three summers later Amazon announced much of what today still powers AWS.
"From experience we knew that the cost of maintaining a reliable, scalable infrastructure in a traditional multi-datacenter model could be as high as 70 percent, both in time and effort, and requires significant investment of intellectual capital to sustain over a longer period of time," wrote Werner Vogels, Amazon's CTO. "The initial thinking was to deliver services that could reduce that cost to 30 percent or less (we now know it can be much less)."
Much of the initial work was done by a small team of engineers in South Africa which built the first version of EC2.
The early EC2 work was based on an Amazon paper that explained how to turn Amazon’s in-house e-comm engine into a standard automated set of cloud services including services.
“At Amazon we had developed unique software and services based on more than a decade of infrastructure work for the evolution of the Amazon E-Commerce Platform. This was dedicated software and operation procedures that drove excellent performance, reliability, operational quality and security all at very large scale. At the same time we had seen that offering programmatic access to the Amazon Catalog and other ecommerce services was driving tremendous unexpected innovation by a very large developer ecosystem,” said Vogels. “The thinking then developed that offering Amazon’s expertise in ultra-scalable system software as primitive infrastructure building blocks delivered through a services interface could trigger whole new world of innovation as developers no longer needed to focus on buying, building and maintaining infrastructure. From experience we knew that the cost of maintaining a reliable, scalable infrastructure in a traditional multi-datacenter model could be as high as 70%, both in time and effort, and requires significant investment of intellectual capital to sustain over a longer period of time. The initial thinking was to deliver services that could reduce that cost to 30% or less (we now know it can be much less). We were also keenly aware that compute utilization in most cases, enterprise as well as startups, is dramatically low (less than 20% and often even lower than 10%) and is often subject to significant periodicity. Providing these services in an on-demand fashion using a utility pricing model had the potential to radically change this. AWS delivered the first storage service (Amazon S3) in the spring of 2006 and compute (Amazon EC2) in the fall of that year. The rest is public history…”
Bezos elaborated on Vogols’ comments in an online posting. “Approximately nine years ago we were wasting a lot of time internally because, to do their jobs, our applications engineers had to have daily detailed conversations with our networking infrastructure engineers. Instead of having this fine-grained coordination about every detail, we wanted the data-center guys to give the apps guys a set of dependable tools, a reliable infrastructure that they could build products on top of,” Bezos said. “The problem was obvious. We didn’t have that infrastructure. So we started building it for our own internal use. Then we realized, “Whoa, everybody who wants to build web-scale applications is going to need this.” We figured with a little bit of extra work we could make it available to everybody. We’re going to make it anyway—let’s sell it.”
Amazon already knew the e-comm engine it used in-house was efficient and economical. It figured it would be a cost effective solution as well. And that has turned out to be true as the TCO and ROI from AWS are both impressive, says IDC in a report sponsored by Amazon.
IDC offers this little AWS history lesson.
“Amazon's focus on driving costs out of the company's large-scale ecommerce operations led Amazon to move toward service orientation and exposing all resources as scalable and consumable services. This movement also ensured that the development culture at Amazon would be aligned with modern development techniques and result in a platform that was flexible, agile, and extensible. This has resulted in a high degree of AWS API envy across the industry. From an AWS customer standpoint, this means it is easier to develop applications using AWS, easier to deploy preexisting applications on AWS, and easier to establish hybrid operations spanning AWS and private datacenters,” IDC wrote the report.
“During the late 1990s and early 2000s, as Amazon was emerging as the world's leading ecommerce company, internal business requirements necessitated that Amazon build out an application infrastructure that would support massive scale and reliability in the following areas: compute, parallel processing, storage, content management, data management (relational and nonrelational databases), transaction processing, messaging, queuing, payments, security, monitoring, and management,” IDC wrote.
IDC findings show that AWS customers save money on IT infrastructure, increased IT productivity and improved end user productivity.
“By accelerating the application development and deployment process, automating application management, and switching to IaaS, IT staff are now 52% more productive, saving nearly $150,000 per application per year. IT staff are thus able to improve support of mission-critical operations,” IDC said.
Meanwhile “End users benefited from fewer service disruptions and quicker recovery, reducing downtime by 72% and saving nearly $32,600 per application per year,” the research house found.