AWS: Running a Hyperscale Data Center

This is a snippet of the transcript, sign up to read more.

To have an idea, you said a third, a third, a third of revenues is going through that region. Could you say what revenue share comes from hyperscale and colocation facilities?

Yes; 60% of our revenue was delivered through services in hyperscale data centers. The financial part of running is paying for people who are not necessarily the easiest to find and train. You've got a whole people cost. You're spending about $1 million in salary and benefits to run an entire region. You also have your facilities, probably your biggest cost, because that's where the power comes from. The more servers you put in, the more power you consume and the higher your cost. Having a full data center will cost a billion dollars a year because of your racks.

This is a snippet of the transcript, sign up to read more.

To have an idea, you said a third, a third, a third of revenues is going through that region. Could you say what revenue share comes from hyperscale and colocation facilities?

Earlier, I mentioned you put about 700 racks in one of these hyperscale data centers. Each rack is $400,000 to $500,000. It is not cheap. When we have semis come up to roll out 12 racks to be installed, you're looking at $4.8 million worth of racks in one semi. That's an enormous cost, your compute. However, we depreciate that over five years. If you've got 700 racks and your average cost of each rack is about $400,000. That's a big cost. The next one is powering them. We look at each rack taking up about 35 megawatts. I'm trying to extrapolate my costs for each rack position here. It was probably about $500 a month per rack, $30 million per year. So $100,000 per megawatt is what we used, so $30 million a year in power for a fully loaded data center.

This is a snippet of the transcript, sign up to read more.

I’m not, go ahead.

PUE is power utilization efficiency, and it's the number one metric data center operators have to report. That means that what percentage of the electrical load is attributable to computing versus other things that could be in there. You’ll see numbers that are less than one, and you rarely see them higher than 1.8; 1.8 is bad, and anything less than one is extraordinarily good. Most companies come in around 1.1, or 1.2, which means the ratio of compute power over the amount of power that’s provisioned. That’s a key metric that you’ll watch. You’ll find that the lower the PUE, the more efficient those data centers are, and they're probably the most successful ones.

This is a snippet of the transcript, sign up to read more.