Today small to medium enterprise businesses (SMB) who have chosen to run their own “in-house” data center have done so using the status quo approach to building a data center. First, guess how much capacity you will need in terms of IT, power, and cooling. Second, build a data center room that supports the maximum final load, which is capital intensive and will likely be underutilized for the life of the data center. Third deploy IT using the siloed approach.

By “siloed” I mean that the data center is composed of specialized software and hardware and supported by several specialized IT personnel with a deep and narrow knowledge of a particular domain. At a high level, these domains include storage, networking, and compute, but can be broken down into things like virtualization experts, and further still into a specific visualization platform like VMWare, or Hyper-V.

This is the status quo method of operating small data centers today because of the way IT hardware and software are offered. The status quo is to buy piecemeal components and integrate them into the business.

– Thinkstock/Brand X Pictures


However, this status quo is being disrupted by hyperconverged IT. Hyperconvergence basically refers to the integration of compute (i.e. processor), storage (i.e. hard disk drive), and networking resources into a single chassis.

Now, imagine you could deploy these chunks of hyperconverged infrastructure in a single rack referred to as a micro data center. Micro data centers are available today, preconfigured and factory tested, in small enclosures containing all the essential components: IT, power supply, cooling, and physical security and monitoring features.

Some of these prefabricated micro data centers are delivered in power ratings of 5kW, ample for the requirements of many SMEs. An additional advantage is that their reliability and modularity also make them ideal building blocks for those wanting to scale up to larger data centers.

Using micro data centers as building blocks can typically remove much of the up-front investment associated with a traditional data center build. They can be deployed in spare space, within existing premises and make use of ‘sunk costs’ already committed to utility power and cooling. They can also provide much more manageable increments of computing resource, allowing increased load demands to be satisfied with much smaller upfront investments.

In many jurisdictions such an approach yields a tax advantage because such small prefabricated systems can be classified as business equipment rather than building improvements and therefore the investment can be depreciated more quickly.

In a recent study, which compared a traditional purpose-built 1MW data center with an equivalent based on 200 5kW prefabricated micro data centers, we found that the micro data centers saved 42% capital expenditure over traditional. Most of the savings was achieved by avoiding the costs associated with both building and fitting out new purpose built premises.

Deploying a system of prefabricated micro data centers over a wide geographical area also provides an excellent level of resilience as the failure of one such micro data center can be absorbed by others on the network taking up its load. Admittedly, this is forward-looking, but not absurd.

Enterprise IT is headed in this direction and many global corporations, including web giants, already use IT to afford them geographical redundancy.

Today there will be occasions in which the status quo, or purpose built approach to data center design and build is preferable to the distributed micro method. If network latency for example, is a vital concern, then having your entire IT infrastructure in the same location may be required. The delays caused by sharing compute and storage between micro data centers far apart from each other could be impractical for certain IT services today. Nonetheless, like many other technical obstacles, advances in IT will overcome this obstacle and will change how we design and build data centers.

Victor Avelar is the director and senior research analyst of the Schneider Electric Data Center Science Center