Data Center Journal

VOLUME 45 | AUGUST 2016

Issue link: https://cp.revolio.com/i/709701

Contents of this Issue

Navigation

Page 8 of 24

6 | THE DATA CENTER JOURNAL www.datacenterjournal.com tion is a requirement for doing business, but it can also offer protections to compa- nies and their customers. e ongoing case of Microso versus the U.S. government is one illustration. An appeals court in July of this year issued a ruling that prevented the government from forcing Microso to hand over emails stored in a data center on foreign soil. At the international level, therefore, decentralization offers some potential privacy advantages, although legislative updates could easily change that situation. But the situations in which decentralization serves regulatory, privacy or related issues may be more mundane as well. Decentralization, if done properly, can also effectively obviate the need for a backup data center, since downtime at one or two facilities doesn't necessarily mean a total loss of service. Other data centers can pick up the slack in these cases, so a company need not maintain an idle or semi-idle facility that simply waits for an outage to strike the main facility. do sMaller, decenTralized daTa cenTers sTand a chance? e idea of decentralized networks as an alternative to the centralization trend is nothing new. Research is underway regarding how such networks can be implemented to overcome the technical challenges of operating in many different locations instead of a single location. For instance, a 2014 paper at the IEEE Interna- tional Conference on Peer-to-Peer Com- puting, entitled "Decentralizing the Cloud: How Can Small Data Centers Cooperate?" addresses some of these very questions. e authors attempt to show how a swarm of smaller data centers can handle work- loads in a unified manner without falling prey to their individual limitations, such as the relatively low capability of a small data center to handle large traffic bursts. Despite the clear conceptual benefits of decentralization, such as service provi- sion closer to users, lower latency, greater overall resiliency and potential regulatory- compliance advantages, economics re- mains a concern. e monetary benefits of owning and operating multiple small data centers—whether instead of or in addition to larger facilities—must outweigh the associated costs. For instance, even a small data center operating in an urban area can incur large real-estate costs compared with one operating at a more distant location. Furthermore, even a highly automated data center will still require human hands from time to time, so operation of many of these decentralized facilities may neces- sitate either frequent travel or a disbursed set of employees to manage certain tasks. Economies of scale may simply fail to ap- ply in this model, whereas they are a major factor driving centralization. So the issue that companies must re- solve is whether dealing with local data lo- cally is worth the savings in long-distance transmission, latency and bandwidth, or whether operating fewer, larger data cen- ters is simply the most economical choice, even when factoring in performance con- siderations. Fortunately or unfortunately, the decision seems to lie almost exclusively with companies. e idealistic vision of an almost totally decentralized network in which large corporations and governments lack the capability to control the flow of data is all but dead. Tablets, smartphones and IoT clients may offer tremendous computing power compared with larger systems from just 10 years ago, but they are ill equipped to realistically serve as nodes in a distributed computing network. Possibly the best candidates for that job, desktop PCs (and, to a lesser extent, note- book PCs), represent a declining market. And as companies abandon in-house IT for the cost benefits of the cloud, even that lesser distribution on a business scale declines. e near future probably won't see unadulterated centralization, but decentralization will play a small role in creating something of a hybrid market. Clearly, decentralization offers technical and other benefits in certain situations, but it also lacks a powerful case to create a strong countertrend. Large companies like Amazon, Google, Facebook and Microso may well dabble with smaller data centers operating at the customer edge, just like they dabble in other technologies and strategies. Unless they pursue it enough to draw the attention of smaller companies, however, it may simply remain a novel curiosity, much like data center containers and immersion cooling—technologies that certainly have a place yet lack the punch to drive broad industry adoption. e situation will ultimately depend on what service consumers demand, as well as what they're willing to pay. conclusions Centralization of IT resources is rais- ing concerns, particularly as mobile and IoT demand increases and as companies increasingly outsource to the cloud. But it carries economic incentives that make any move back toward decentralization a tough proposition. Deployment of smaller "edge" data centers closer to customers most certainly offers a number of techni- cal benefits, including amelioration of bandwidth limitations as well as lower latency. For the many tasks involving data that need not go to distant locations, these decentralized facilities seem like a great way to meet demand while providing peripheral benefits such as greater network resiliency and possibly regulatory-compli- ance advantages. Nevertheless, the trend toward centralization is a huge train with lots of momentum, so the technical landscape will need to change before decentralization becomes a real movement. e IoT, in the event that predictions claiming gargantuan numbers of connected devices on the ho- rizon are remotely accurate, may produce conditions that drive greater decentraliza- tion. In the meantime, big IT players will probably determine how the situation un- folds over the next decade. With Moore's Law failing and political tensions growing, however, the conditions are becoming more fluid and the character of the data center industry more difficult to predict. Given that centralization and decentraliza- tion are two extremes on a pendulum and that centralization seems to be at its peak, it's mostly a matter of how long it will take the industry to change course. n

Articles in this issue

Links on this page

Archives of this issue

view archives of Data Center Journal - VOLUME 45 | AUGUST 2016