Saturday Night Live used to run a skit entitled “Who is Muy Macho?” in which older celebrities were compared by the host to determine who was the more macho of the two. If they were to use the same skit today they could call it, “Who is Muy Modular?” How better to describe the current state of affairs in our industry where everything from a 12’x40’ container to a 500,000 square foot building can be promoted as offering customers the market’s most modular alternative. A 21st century Theater of the Absurd equivalent.
Since there are no standards for modular components, or even a single accepted definition of what defines a “modular” data center, the term is tossed about so casually that, from a customer perspective, it is virtually worthless. Although it implies transportability and flexibility the majority of “modular” offerings either reside within a single vendor provided shell or require the customer to provide their own. Neither of these alternatives is terribly flexible in terms of configuration options and they certainly don’t unshackle the data center from a fixed geographic location. In terms of data centers, modular could be more accurately defined as meaning, “located in a large building in NY/NJ, Northern Virginia, Chicago, Dallas, San Francisco or Phoenix”.
Even the appeal of incremental expansion is diluted with most modular offerings. For example, in many monolithic building-based data centers the finite space available within the facility requires that customers lease the additional expansion space they will require up front. As a result of this need to pre-lease space the link between data center spending and expansion is broken and the value of incremental growth is lost. In the case of container-centric solutions, the relationship between spending and growth remains more or less intact but often at the cost of geographic flexibility and reliability due to their use of single “daisy chained” backplanes. In other words, should a Kansas-based company’s only “modular” alternative be to add another container in Chicago or Dallas?
The pretense of using “modular” as a method of data center classification is further eroded when we consider that we are categorizing based on a mode of construction and not the end product. We need to look no further than the current confusion over whether pre-fabricated units are synonymous with modular (they are not). The point here is that once again we are arguing semantics and not substance. All this confusion points to the need for a new descriptive classification for data centers focused on the end product and not the means used to reach it.
When we examine the issue from a customer perspective we need to look at the evolving state of their requirements. The increasing need for data center space is generating a new requirement for ubiquity in terms of solutions. The dispersed geographic requirements for enterprise level firms and cloud providers have created a need for facilities that “go to the customer” rather than the “build it and they will come” model preferred by the majority of today’s providers. Despite possessing a perceived level of mobility, containers cannot fulfill these new requirements for geographic independence due to their non-hardened designs that make them susceptible to environmental factors like high winds. Putting $20M of IT gear in “data center” that can become airborne when the wind blows can lead to a lot of sleepless nights.
As customers continue to demand more data center geographic flexibility they also are calling for more control. This need for enhanced control necessitates that providers make BMS and DCiM integration a central component of their offerings. Previously, many providers viewed this systems linkage as an afterthought compelling their customers to “live” with what was available. A greater level of sophistication in terms of overall infrastructure and computing equipment monitoring and administration is rapidly becoming the rule and not the exception for customers. This desire for greater control also extends to the physical facility itself. If customers now require their data centers where they want them, it is not unreasonable to assume that they will no longer accept the idea of sharing a facility with one or more additional customers.
If we look at these requirements in the aggregate it becomes apparent that the logical point of differentiation for data centers isn’t their degree of modularity but whether or not they can provide the complete single source solution that customers desire. These new customer requirements necessitate the need for the “productization” of data centers. Under this structure all of the necessary elements (geographic independence, hardened design, and enhanced customer control) of the data center must be “packaged” within a single product. This new structure would simplify the technological review process through the inclusion of check box items such as Uptime Tier III certification, full Level 5 commissioning, high-density support, and PUE thresholds. Modular design may be the underlying element of these new offerings but they are not their defining characteristic. I would suggest that this new breed of comprehensive data center products be defined as “Standalone Data Centers” to enable customers to evaluate competing offerings based on their actual capabilities rather than their proximity to an ill-defined architectural definition.