Trying to assess just how enormous Amazon Web Services is has become a sort of parlor game among techies. Counting servers is as good a way as any to get a grip on its size and the latest to take a stab at that is Netcraft, which pegs the numbers of AWS web-facing servers at 158,000, up from 118,000 such servers in September, 2012. (Hat tip to Data Center Knowledge for pointing out this interesting research.)
In other exploratory forays, just over a year ago an Accenture analyst, using his own methodology, put the total count of AWS servers at 450,000 or so, but that figure included boxes in the guts of the system, not just web-facing servers. In April 2012, researcher Deepfield Networks estimated that a full 1/3 of all internet users touched an AWS-based server at least once a day and that 1 percent of all consumer traffic moved either into or out of the AWS cloud.
Figuring out just what goes on within AWS often feels like the blind men and the elephant parable, with each researcher gleaning some information about the whole but none able to provide the whole picture. But the reason people keep trying is that Amazon itself does not provide details on just how much gear it uses. Only that it adds more and more of it all the time. You’re welcome to ask Amazon CTO Werner Vogels about all this at Structure later this month.
The sheer scope of AWS is freaky — and the fact that most people estimate that AWS is bigger than a $2 billion-a-year business now — is one reason every tech vendor HP,Oracle, IBM, VMware, Microsoft, and Google are fighting for their piece of the public cloud pie.
The more traffic AWS handles, the more it learns about user behavior patterns and the more it can apply that wisdom to tweaking its infrastructure design to reducing cost while also (shocker!) making money on it. Talk about your virtuous cycles.