As Microsoft gets closer to bringing its Chicago Data Center container on line, and deploys more server containers than are used by they rest of world. I started speculating on possible outcome as almost all the large server OEMs prepare containers of servers. Here are a few thoughts on the subject.
- In the past you bought a server, installed it in a rack. You could buy a rack full of servers. But, now you can buy a container full of servers all configure with the exact amount of power and cooling infrastructure required for the servers in the container.
- A large portion of the power and cooling infrastructure have been packaged with the server container. The server power and cooling infrastructure is now encapsulated as part of the container.
- It should be much easier to deliver a power and cooling optimized environment for a container, than a typical white space. Vendors can make the economic trade-offs of performance in all kinds areas to squeeze the best use of power and cooling energy.
- Designs can be optimized for different costs data center CAPex costs and OPex costs. And, could be part of the selection criteria. Makes sense to pay more for a higher efficiency container if you higher power costs.
- Microsoft’s strategy of having multiple servers OEMs bid on containers plays to this advantage. And, is a disadvantage for Google as they would have built the container all on their own and kept the IP. Which could explain Google’s decisions not to use containers.
- Google thinks they are smarter than the server OEMs.
- Microsoft thinks the server OEM ecosystem does a better job than they can when it comes to designing server HW, including the right power and cooling infrastructure.
- As power and cooling CAPex and OPex continue to rise as a % of data center costs, it makes sense to focus on how to reduce CAPex and OPex. Compute will be a commodity, and costs need to be lowered.
- Does this sound like Cloud Computing’s future?