Issue link: https://cp.revolio.com/i/709701
20 | THE DATA CENTER JOURNAL www.datacenterjournal.com infrastructure in a factory-built module. ese data centers may, for instance, come in a shipping-container-style "box" and are suited to remote areas. Some vendors offer rack-level modules that may be suitable for deployment at a company's satellite loca- tion or a colocation facility. In the latter case, little or no employee presence may be necessary, allowing the company to bring services closer to users without incurring the costs of remotely stationed personnel or extensive travel. e availability for the aggregate of micro data centers will also depend on a number of factors. An outage at one may only affect local users, but for those users, it would appear as a complete service outage unless a backup picks up the slack. at strategy might involve failover to a central facility, another micro data center (or some combination of them) or both. Whatever the case, individual availability ratings become much less of a concern for micro data centers (assuming they are deployed up to industry standards) than for a small number of large data centers. Nevertheless, specific requirements will depend on the service that the provider is offering and how critical it is to the busi- ness and to customers. disBursing The cosT A major factor driving the centraliza- tion of data centers is cost: specifically, the economies of scale that come from large equipment purchases and amortization of the expenses across many customers. is model underlies the cloud, pushing a growing number of companies away from the expenses of building and operating their own private data centers. Micro data centers and the decen- tralization they represent would seem to be at odds with these cost advantages. Nevertheless, the modularity of these data centers can enable cost savings through a different economy of scale: mass manufac- ture. Prefabricated units for deployment in a number of locations can certainly deliver savings relative to custom design and con- struction of each unit individually. Moreover, going where the custom- ers are may be an expensive proposition— particularly in urban areas. Granted, small data centers require less space, but when deployed across many metropolitan areas the costs can add up. Urban colocation may be a better option than, for instance, a self-contained "data center in a box," but it still commands a premium compared with the rural locations in which many big- name companies have built large facilities. But even if the aggregate infrastruc- ture of micro data centers is more costly than that of a central facility, the opera- tional benefits may be worth it. As noted above, decentralization can ease availabil- ity constraints, and the lower latency and greater bandwidth that these smaller units enable can increase value for customers, potentially increasing revenue. A success- ful micro data center deployment, how- ever, requires careful planning with regard to management of remote sites, strategies for addressing outages and so on; unfortu- nately, the situation is more complex than simply moving racks out toward custom- ers. Like most projects, however, identify- ing the best approach requires a careful analysis of the costs of each one and the likely market benefits. planning for availaBiliTy A number of market factors are driv- ing interest in micro data centers designed to deliver services locally at the customer edge rather than over long distances. But the strategy for such a deployment will differ considerably from that of the traditional central data center. A major difference is availability: no longer is a single number (three or four nines, for instance) truly descriptive when many micro data centers are in play. If these small facilities act as a mutually supportive network, the availability of the whole can be much greater than that of each one individually. But garnering the benefits of this approach requires careful planning, since it may come at the expense of some cost savings associated with centralization. Customer demand for lower latency and greater bandwidth to serve mobile and IoT applications are pushing companies in this direction nevertheless. Decentralization is still at odds with the larger market trend toward fewer data centers in the hands of a shrinking number of companies, thanks in part to growing reliance on the cloud. But the history of computing has shown some back and forth, and the technological landscape may be reaching a point where geographical decentralization becomes economical, even if the pool of companies that own most of the resources remains small. If that happens, the focus on high- availability data centers may ease as pro- viders deploy more edge networks—even if centrally managed—rather than a few large hubs. n Decentralization is still at odds with the larger market trend toward fewer data centers in the hands of a shrinking number of companies, thanks in part to growing reliance on the cloud. But the history of computing has shown some back and forth, and the technological landscape may be reaching a point where geographical decentralization becomes economical, even if the pool of companies that own most of the resources remains small.