Velocity 2010: James Hamilton, "Datacenter Infrastructure Innovation"

According to James Hamilton who is a distinguished engineer at AWS no: link and link And you can watch him say it too: ▶ this video

The comment in general may be true but I don't see electricity costs being "much greater than CPU purchase costs over the lifespan" at least not for a data center operator. The AWS M4 EC2 instance uses "2.3 GHz Intel Xeon® E5-2686 v4 (Broadwell) processors or 2.4 GHz Intel Xeon® E5-2676 v3 (Haswell) processors" [0] so about 4-6 years old and may approach 7 years. A comparable processor Intel® Xeon® Processor E5-2670 v3 costs about $ 1600 [1] in bulk. Average US home electricity price is ~ 0.13 $/month [2], this would probably be much lower for a data center operator. The mentioned processors probably consume ~ 150 W at full load, which will probably not be the case most of the time. Anyway, let's see: CPU cost over its life time may be on the order of 1600 USD, which is quite cheap as today's top processors cost usually between 4000 and 8000 USD. [3] The electricity cost would probably be: 24 hours * 365 days * 7 years * 0.13 $/kWh * 0.15 kW ~ 1200 USD which seems about right. We can observe the electricity cost per CPU to be significantly lower than the purchase price of a rather cheap processor even assuming very generous CPU load and electricity price, while assuming rather low CPU cost. This of course doesn't include the mainboard, RAM, NIC, power conversion, cooling and other hardware that is needed to operate a modern data center. On the other hand, the cost of this hardware wasn't included in the calculation either. Of course, I am not the only one to do the calculatio

I may be missing something, but that post doesn't address the expense of power, it talks about the relative consumption of networking equipment compared to the rest of the facility. Power is expensive, and James Hamilton generally spends lots of time talking about it, his Velocity keynote ( ▶ this video ) about datacenter innovations focused primarily on power and cooling (heat being a biproduct of power consumption).

That split of where the money goes comes directly from calculations and a spreadsheet that James Hamilton has developed. You should check out his Velocity talk for more (about 5 minutes in) ▶ this video The numbers on the slides modify the assumptions on the spreadsheet at link