Amazon Web Services (AWS) Hardware
Unlike most of its competitors, Amazon builds its hardware infrastructure out of merchandise components. Goods, in this case, refers to the use of equipment from less well-known manufacturers who charge less than their brand competitors. For components for which there are no merchandise offerings, Amazon (known as a fierce negotiator) gets very low prices.
On the hardware side of the AWS offering, Amazon’s approach is clear: Buy equipment at the cheapest possible price. But wait, you might say, wouldn’t the commodity approach result in a less reliable infrastructure? After all, providers of branded hardware assure that one of the advantages of paying premium prices is that you get quality equipment.
Okay . . . Yes and no. It may be true that premium priced equipment (traditionally called enterprise equipment due to the assumption that larger organizations require more reliability and are willing to pay more for it) are more reliable in the apples-to-apples comparison. That is, an enterprise-grade server lasts longer and experiences fewer outages than its commodity-grade counterpart.
The problem, in Amazon’s view, is how much more reliable Enterprise equipment is than the commodity version, and how much that improved reliability is worth. In other words, you need to know the cost-benefit ratio of the enterprise versus the goods.
Making this assessment more difficult is a basic fact: at the scale at which Amazon operates (remember, it has nearly half a million servers running on its AWS), equipment—no matter who provides it—fails all the time.
If you’re a cloud provider with an infrastructure the size of Amazon, you have to assume, for every type of device you use, an endless round of malfunctioning drives, fried motherboards, network switches dropping packets, etc.
So, even if you buy the highest quality and most expensive hardware available, you will end up (if you are lucky enough to switch to a very large cloud computing provider like Amazon) with an unreliable infrastructure.
In other words, on a very large scale, even highly reliable individual components can still lead to an unreliable overall infrastructure due to component failure, which is as rare as the failure of a particular piece of equipment.
The scale with which Amazon operates affects other aspects of its hardware infrastructure as well. Besides components such as servers, networking, and storage, data centers also contain power supplies, cooling, generators, and backup batteries. Depending on the specific component, Amazon may have to use equipment specifically designed to operate at the required scale.
Think of AWS’s hardware infrastructure this way: If you have to design and operate data centers to handle massive scale and in a way that aligns with the company’s mandate to operate inexpensively, you’ll probably end up with a solution much like Amazon’s. You can use commodity computing equipment whenever possible, unwind prices when you can’t get supply of commodities, and custom design equipment to run your extraordinarily large scale operations.