Cutting through the Myth, what is a Modular Data Center

I've been writing about modular/container data centers for a while and I stopped trying to post on all the different news.  How bad is it?  How about 35 suppliers.  See what DatacenterKnowledge posts.

A Shift in the Modular Conversation

This year’s Uptime Symposium reflected an ongoing shift in the industry discussion of modular designs, with a growing focus on economics and other factors directly tied to purchasing decisions. The event featured three keynotes and an entire content track on the economics of modular data centers, along with a modular data center campus in the expo hall featuring offerings from four providers.

“We’re tracking 35 suppliers,” said Lawrence. “They think this is a market worth chasing. It’s undeniable that it is a very effective means of deploying data center space for some users. The adoption won’t be linear, and won’t be universal.”

Compass Data Centers is trying to educate.

“Modular—Composed of standardized units or sections for easy construction or flexible arrangement”—Random House American Dictionary 

But, I think this still goes over many heads.  Let's try this for an explanation.  

All this stuff about Modular is just the idea of interchangeable parts at data center scale.  I am going to scrape parts from the wikipedia post in interchangeable to illustration a few points

Ok first what are interchangeable parts?

Interchangeable parts are parts that are, for practical purposes, identical. They are made to specifications that ensure that they are so nearly identical that they will fit into any device of the same type. One such part can freely replace another, without any custom fitting (such as filing). This interchangeability allows easy assembly of new devices, and easier repair of existing devices, while minimizing both the time and skill required of the person doing the assembly or repair.

Ford assembly line, 1913. The magneto was the first to be assembled.

The concept of interchangeability was crucial to the introduction of the assembly line at the beginning of the 20th century, and has become a ubiquitous element of modern manufacturing.

Yep. I want my data center to built on interchangeable parts that does all of the above.  Much clearer explanation than I want modular.

What was the technology behind interchangeable parts?

Interchangeability of parts was achieved by combining a number of innovations and improvements in machining operations and the invention of several machine tools, such as the slide rest lathescrew-cutting latheturret lathemilling machine and metal planer. Additional innovations included jigs for guiding the machine tools, fixtures for holding the workpiece in the proper position, and blocks and gauges to check the accuracy of the finished parts.[1] Electrification allowed individual machine tools to be powered by electric motors, eliminating line shaft drives from steam engines or water power and allowing higher speeds, making modern large scale manufacturing possible.[2] Modern machines tools often have numerical control (NC) which evolved into CNC (computerized numeric control) when microprocessors became available.

Cutting tools made of high speed steel allowed steel rather than wrought iron to be used for parts.[3][4] The ability to machine hardened parts eliminated the problem of warping and dimensional changes associated with heat treatment hardening of parts after machining.[5] Modern cutting edges also use materials such as tungsten carbide. Other innovations were drop forging and stamped steel parts, which reduced or eliminated the amount of machining in the United States.

Being able to put the pieces all together just like above is the goal.

Now somehow with the momentum of Modular Data Centers you are lead to think that the past data centers were custom built.  Like the way guns were made before the 18th century.

Before the 18th century, devices such as guns were made one at a time by gunsmiths, and each gun was unique. If one single component of a weapon needed a replacement, the entire weapon either had to be sent to an expert gunsmith for custom repairs, or discarded and replaced by another weapon. During the 18th and early 19th centuries, the idea of replacing these methods with a system of interchangeable manufacture was gradually developed.[6][7] The development took decades and involved many people.[6][7]

This is of course ridiculous to think that the last 5 years of data centers are custom built.  

The data centers built 25 years were custom, but now everyone uses interchangeable parts. 

Eli Whitney showed how interchangeable parts could revolutionize gun manufacuring.

Eli Whitney and an early attempt

In the US, Eli Whitney saw the potential benefit of developing "interchangeable parts" for the firearms of the United States military. In July 1801 he built ten guns, all containing the same exact parts and mechanisms, then disassembled them before the United States Congress. He placed the parts in a mixed pile and, with help, reassembled all of the weapons right in front of Congress, much like Blanc had done some years before.[8]

Doesn't the sound like presentations and videos that have been shown that all the pieces go together.  

Eli Whitney sold Congress.

The Congress was captivated and ordered a standard for all United States equipment. Interchangeable parts removed problems concerning the inability to consistently produce new parts for old equipment without significant hand finishing that had plagued the era of unique weapons and equipment. If one weapon part failed, another could be ordered, and the weapon wouldn't have to be discarded. The catch was that the Whitney's guns were costly and handmade by skilled workmen.

Note this last part. "costly and handmade by skilled workmen."  

Eli Whitney overpromised and under delivered.

Whitney was never able to design a manufacturing process capable of producing guns with interchangeable parts. Fitch (1882:4)[6] credited Whitney with successfully executing a firearms contract with interchangeable parts using the American System, but historians Merritt Roe Smith and Robert B. Gordon have since determined that Whitney never achieved interchangeable parts manufacturing. His family's arms company, however, did so after his death.

Now welcome to the part of Modular Data Center construction that almost no one talks about, especially in public.  The projects that don't work.

I was sitting next to a Telco data center guy at dinner on Tuesday night. And, modular data centers came up and I said yeh, those overhyped heavily marketed containers.

He pulls out his phone shows pictures of his containers he would like to sell.  I wonder if he can get them  listed on eBay or Craigslist?  He hates his containers.  

They are expensive to operate and hard to repair.

So, does this mean modular data centers will not work.  No.  

But, the probability of users being able to pick the right solution with 35 companies and heavy marketing, the odds are not good.