Modular Design for Small Data Centers

August 6, 2012 5 Comments »

Modularity in Design

Modular data center design is a hot topic, and as companies struggle to meet their IT needs and satisfy stringent budget limitations in a slow economy, any design approach that promises to reduce capital and operational expenses and simplify builds and expansions is fodder for consideration. Wholesale and colocation data center providers have become experts in modular design owing to the market they serve, where “right-sizing” of infrastructure—yet the ability to expand when needed—is a critical part of success. But what about companies looking to build small data centers? By looking at the practices of data center service providers, these companies can learn lessons in modularity that can yield tremendous benefits.

The First Lesson: Modularity Does Not Necessarily Mean Containers

For whatever reason, modularity has become nearly synonymous with ISO-standard containerized data centers—but that is a very narrow interpretation of modularity. Although data center containers are one particular form of modular design, they are far from the only such manifestation. Modular design, although it might involve large, mobile data-centers-in-a-box, can instead focus on specific systems (IT or facilities) in a traditional data center, and the scale can range from large to small. Rather than being isolated to movable containers, modularity is a means of designing and building data centers in a manner that meets or only slightly exceeds current requirements while enabling expansion via relatively simple addition of new infrastructure (as opposed to retrofitting existing infrastructure or constructing entirely new facilities).

Data Center Containers: The “Paragon” of Modularity May Not Be for You

For a company looking to build a small data center, the promises of containerized data centers are tempting. Fast, “plug-and-play” deployment. Lower cost. No need to build a traditional structure (just drop the containers in your company’s parking lot!). Mobility. But the benefits that are often ascribed to containers may not materialize, particularly for a small company (or any company building a small data center facility). According to an EC&M article (“Modular Data Center Design Trends”), these benefits could well be overstated: “the promise of a lower cost data center hasn’t always come true, especially in comparison with small, early phase traditional data centers. In fact, the cost of the container data centers often reaches the price of a Tier 4-classified center but without the elements offered by that level of data center.” Furthermore, beyond potential issues with reliability and vendor lock-in, “the speed of deployment, described as plug-and-play, is oversimplified. Container designs still require a chilled water supply and a secure site, and, in some cases, applications for building permits could delay deployment for months.”

Thus, companies building small data centers shouldn’t assume that containers are equivalent to modularity—nor should they dismiss traditional data construction out of hand. If cloud computing and colocation don’t fit the bill, data center containers are not the only choice. Indeed, the container market is actually still quite small, and the lack of broad deployment (meaning a lack of plenty of examples to follow) mean this route may be less than ideal for a company that wants a solid, reliable and inexpensive as possible facility.

Applying Modularity on a Small Scale

Conceptually, the modular approach is particularly ideal for companies building smaller data centers. Demand for IT resources is generally increasing, and most companies can and should expect their data center needs to increase over time. But building to cover anticipated demand, say, five years down the road invariably requires guesswork and results in greater near-term capital costs. Equipment that is purchased but unused must be stored and possibly maintained; if connected, it can consume power, increasing operational expenses even though no commensurate service is being supplied.

This reflects the situation for colocation/wholesale data center providers. To maximize profitability, these companies must provide enough infrastructure (cooling, power distribution, network connectivity and so on) to meet customer demand, but going overboard can be costly. Unused equipment has a shelf life. Imagine, for example, buying a car now to meet the anticipated need of a growing family 5 to 10 years in the future. Over that time, the car only loses value—and it is still subject to certain kinds of wear that could affect its functioning. The same is basically true for data center infrastructure. Buying servers now, for instance, when they might not be needed for several years has several downsides:

  • Capital, which could be used elsewhere, is spent on equipment that performs no useful function for some time.
  • Equipment loses value over time, even if it is unused.
  • Even unused equipment may require maintenance.
  • Exploiting new technologies (such as advances in processor manufacturing technology) means buying new equipment to replace old equipment that may never have been used.
  • Precious storage space—or possibly data center floor space—is consumed by idle equipment.
  • Power is consumed by idle equipment if it is connected and turned on.

Similar problems plague other types of equipment that are purchased to “future-proof” the data center.

Modularity isn’t necessarily opposite to future-proofing, but it plans ahead in a way that meets current requirements while minimizing costs and enabling agile expansion when needed. It can be an effective approach regardless of scale, meaning that the design tactics of a wholesale or colocation provider can be applied successfully by a company building a small data center as well.

Modularity Applies to More Than Just IT

IT is the natural first consideration when one thinks of applying modularity to data center construction. As discussed above, why buy a bunch of unneeded servers and other IT equipment? Instead, a company—whether building a small or large data center—could add new server racks when usage approaches capacity. These racks might even be preassembled: they can be shipped fully loaded with equipment, ready to roll into place and connect. But IT is not the only application of modularity.

Cooling infrastructure is a major power drain for most data centers. The average PUE of data centers is estimated to be around 2.0, meaning that only half of all power consumed actually goes to the servers and other IT equipment. The other half is consumed by infrastructure: power distribution and cooling, primarily (although some is consumed by lighting, security systems and so forth). Required cooling infrastructure is proportional to the power consumed by the IT equipment. Thus, deploying more cooling capacity than the data center needs makes little sense if it can be added modularly as needed. A colocation provider with unused space, for example, would want to avoid maintaining unused equipment, in addition to avoiding the unnecessary capital expense, but would want to be able to quickly expand capacity to meet demand from new customers. The same principle applies to companies building smaller data centers.

One particular application of modularity in this context is row-based cooling. A whole-room approach to cooling may be excessive, particularly if not all the floor space in the data center facility is being used. By focusing on rows instead of the entire room, cooling units can be positioned to yield maximum efficiency. Furthermore, as more rows are added in response to growing demand for IT services, more cooling units can be added to support the new equipment.

Power distribution is the other major area benefitting from a modular design approach. From uninterruptible power supplies (UPSs) to power distribution units and backup generators, adding new equipment as needed rather than to cover anticipated future needs provides a number of benefits. For instance, by not overprovisioning UPS capacity a data center can reduce power wasted by these units. A modular solution, however, enables scaling of power capacity when demand requires it. The main cost of the modular approach is more-careful planning to facilitate future expansion when demand approaches or meets capacity.

Containers Not Out of the Question

Although modularity does not necessarily imply a containerized approach, data center containers are not necessarily the wrong choice for a company building a small data center. If planned and executed properly, containers could be an adequate solution, although they do not provide modularity in the same way as would be applied in a traditional facility. (Expansion in the containerized approach is in fairly large units; in a brick-and-mortar facility, it can be in smaller increments.) A company might choose a halfway measure: for instance, it might build a fairly standard data center, deploying IT equipment inside the building, but use containers for cooling infrastructure. In either case, each company must carefully evaluate its options and compare with the advantages and disadvantages of each approach. Not everything that works for a large company with a mega data center will work for a small company.

Conclusions

The main take away for companies building small data centers is that modular design strategy does not necessarily equate with data center containers. Don’t dismiss modularity simply because it brings to mind industrial-looking ISO containers hauled in by tractor trailers. Modularity is far broader a concept, and many companies—particularly some colocation and wholesale data center service providers—have employed it extensively to maximize their profitability. Modularity combines a right-sizing approach with scalability, allowing a company to focus on current IT demand while maintaining the ability to expand quickly and without redesign of an existing facility or construction of a new facility. The benefits of modularity are primarily in the area of capital expenses. Unneeded equipment isn’t purchased for storage until it is needed (an approach that diverts capital from other uses and ultimately results in lost value). Furthermore, storage and data center space is also saved. But in some circumstances, even operating expenses are reduced—particularly when unneeded servers are not deployed and when excess UPS capacity is avoided, thus increasing operating efficiency.

Vendors offer numerous products focusing on meeting the needs of companies targeting a modular approach to building their data centers. For inspiration in their efforts, companies designing small facilities can look to colocation and wholesale providers, many of whose business models have encouraged them to employ modularity to meet customer demand as efficiently as possible.

About Jeff Clark

Jeff Clark is editor for the Data Center Journal. He holds a bachelor’s degree in physics from the University of Richmond, as well as master’s and doctorate degrees in electrical engineering from Virginia Tech. An author and aspiring renaissance man, his interests range from quantum mechanics and processor technology to drawing and philosophy.

Pin It

5 Comments

Add Comment Register



Leave a Reply