Clearing the Air on Cloud Computing

© 2010 John Leslie King

EDUCAUSE Review, vol. 45, no. 3 (May/June 2010): 64-65

Gartner's "Hype Cycle Special Report" said this about cloud computing in 2009: "The levels of hype around cloud computing in the IT industry are deafening, with every vendor expounding its cloud strategy and variations, such as private cloud computing and hybrid approaches, compounding the hype."¹ Gartner puts cloud computing at the "Peak of Inflated Expectations," about to slide down into the "Trough of Disillusionment." Yet higher education is anything but disillusioned: many have bought into Nicholas Carr's vision of information technology in the early twenty-first century as following the lead of electric power in the early twentieth century, with big "generators" providing service through a "grid."²

Cloud computing might or might not be a way of the future, but in any case, it will be understood only in retrospect. IT history is full of dashed hopes and big surprises. Magnetic bubble memory and the Internet broke on the scene in the 1970s. Both were expected to change the world: magnetic bubble memory didn't, and the Internet did. Cloud computing is yet another story in a long-running argument about whether to create large, consolidated service centers and link them to users through data networking or whether to distribute computing resources among users. There is a trade-off: consolidated resources are often more efficient, and distributed resources are often more effective in meeting users' needs.³ Cloud computing is just another episode in a fifty-year-old debate.

Cloud computing does bring some interesting twists to the debate, however. Anyone considering cloud computing should start with the excellent analysis provided by several University of California-Berkeley researchers.⁴ They define cloud computing as the union of Software as a Service (SaaS) and the "public cloud" of pay-as-you-go services that are available to anyone who wants to buy them. (The researchers do not consider strategies that are wholly internal to organizations to be cloud computing.) Cloud computing can offer advantages due to economies of scale in provision of cycles and storage, especially in an era in which electric power costs have become an important component of large-scale computing centers.⁵ There are also potential savings in system administration if the service provider is able to accommodate the activities of many users through automated resource-assignment without service degradation. This is a return to the time-sharing of the mainframe era — but implemented on parallel computing architectures using much more elaborate network technology.

Whether customers can benefit from cloud computing depends largely on the kind of IT work they do. The Berkeley report notes that cloud computing offers potential advantages for mobile interactive applications, batch processing of parallelizable jobs, analytical processing of large data sets, back-end execution of computationally intensive desktop-application work, and other applications that do not require very fast turnaround. (Long network paths introduce latency that can kill some applications, such as fast transaction processing.) This brings to mind the trade-offs of "batch" vs. "interactive" computing and on-site vs. RJE (remote job entry) from back in the day. Any institution considering cloud computing must check carefully to be sure that the specific tasks to be done can be done effectively using a cloud model and that the benefits of doing the work using a cloud model exceed the costs over time. As with nearly everything else in the computing world, details matter.

There are reasons to be skeptical about the benefits promised from cloud computing. Everyone in the IT business knows that things almost never work out exactly as hoped, and the list of IT oracles brought down by hubris is long. (That is the underlying insight of the Gartner "Hype Cycle.") In particular, there are three unknowns that could prove to be difficult, if not fatal, for cloud computing.

The first is the possibility that pressure from cloud computing will drive existing alternatives to improve their efficiency. Steamships replaced sailing ships between 1830 and 1930, but that transition took a century! As the early steamship industry gained ground, the advanced sailing ship industry improved its performance to match. Cloud computing might appear advantageous now, but the current ways of doing things could improve. That's what happened to bubble memory: the disk drive industry surged ahead with declining prices and improving performance, leaving bubble memory behind. Bubble memory did not actually die: similar "non-volatile memory" is alive and well in USB thumb drives and "diskless" laptops like the MacBook Air with the solid-state drive. But spinning disk technology improved faster than bubble memory, and the predictions of victory by bubble memory proved wrong.

Another unknown is how factors taken for granted in conventional systems will fare under cloud computing. The most significant of these is security. A fascinating paper produced by the Cloud Security Alliance, a group of IT security professionals evaluating security issues in cloud computing, lists more than two hundred specific recommendations to be considered before going into the cloud. The concerns raised by these security experts might all be resolved in favor of cloud computing over time, but it is still early. The authors of the security report note, "The path to secure cloud computing is surely a long one, requiring the participation of a broad set of stakeholders on a global basis."⁶ This does not sound like a guarantee that cloud computing will meet the needs of any organization, much less all organizations.

Finally, cloud computing raises important questions about the role of political and legal jurisdictions in promulgating and enforcing laws, regulations, rules, and social conventions related to IT activities. An essential feature of cloud computing is the separation of use and service provisions, with the likelihood that users will be in one political/legal jurisdiction while their data and processing activities will be in another. Cloud computing will almost certainly exacerbate the problems currently seen in transborder data flow, enforcement of laws against political content, or pornography on the Internet. It is already difficult to understand the full implications for FERPA or HIPAA compliance when using remote services to handle student or patient records. Such issues can be sorted out over time, but it is not yet clear whether and how they will be sorted out.

Given that much of the argument in favor of cloud computing rests on reasoning by analogy — that IT services are evolving toward a "utility" model similar to electricity — it makes sense to ask about the quality of the analogy. In fact, the analogy is seriously flawed. The utility model is far from stable, especially with regard to electricity. There are huge shortcomings with the current electric "grid." Long-range electricity transmission results in much loss of energy, and the grid as it exists cannot support distributed energy generation from wind, solar, micro-turbine generation, and other modern technologies. The current effort to apply information technology to electricity in the form of the "smart grid" suggests that electricity should be more like information technology, not the other way around.

Information technology and electricity are superficially analogous in that they both involve electrons and networks, but at a deeper level they are quite different. Electricity provides electromotive force, resistance, and ionization for running motors, heating, lighting, and some industrial processes. This makes electric power complementary in many applications, but in few cases beyond the provision of electric power is electricity the business itself. In contrast, information technology can be used to represent almost anything. In many important fields of commerce (e.g., financial services and insurance), entertainment (e.g., digital games), and research (e.g., computational sciences), information technology is the business itself. The analogy between electric power and information technology is interesting, but it is superficial, and taken too far, it is naïve and misleading.

The next few years of cloud computing are likely to be characterized by a phenomenon well-known in the study of technological innovation. This is "anticipatory retardation," sometimes called the "penguin effect."⁷ It occurs when the potential adopters of an innovation want to move forward but fear of consequences causes them to hold back until they see what happens after the "first movers" take the plunge. (Penguins at the edge of a hole in the ice will wait for another penguin to dive in first, so that they can see if a shark or leopard seal is waiting.) Higher education institutions that want to explore the potential of cloud computing will probably wait until they see that other colleges and universities have made the move and are doing well before they commit.

The cloud computing debate has taken on features of an "inevitability argument," in which proponents suggest that a particular outcome is unavoidable and so people might as well get used to it. True, cloud computing might prove to be a good idea. But that does not mean it is inevitable. Besides, if it is inevitable, what's the rush?

1. "Gartner's 2009 Hype Cycle Special Report Evaluates Maturity of 1,650 Technologies," press release, August 11, 2009, <http://www.gartner.com/it/page.jsp?id=1124212>.
2. See Nicholas G. Carr, The Big Switch: Rewiring the World, from Edison to Google (New York: W. W. Norton and Company, 2008).
3. J. L. King, "Centralized vs. Decentralized Computing: Organizational Considerations and Management Options," ACM Computing Surveys, vol. 15, no. 4 (1983), pp. 319-45.
4. Michael Armbrust, Armando Fox, Rean Griffith, Anthony D. Joseph, Randy H. Katz, Andrew Konwinski, Gunho Lee, David A. Patterson, Ariel Rabkin, Ion Stoica, and Matei Zaharia, "Above the Clouds: A Berkeley View of Cloud Computing," EECS Department, University of California, Berkeley, Technical Report No. UCB/EECS-2009-28, February 10, 2009, <http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.pdf>.
5. A table in the Berkeley report notes a five-fold difference in power costs between Idaho hydroelectric generator sites and Hawaii sites burning imported generator fuel (3.6¢ vs. 18¢ per kilowatt hour). "Above the Clouds," p. 6, table 3.
6. Cloud Security Alliance, "Security Guidance for Critical Areas of Focus in Cloud Computing," version 2.1, December 2009, p. 4, <http://www.cloudsecurityalliance.org/csaguide.pdf>.
7. See W. Fellner, "The Influence of Market Structure on Technological Progress," Quarterly Journal of Economics, 1951, pp. 556-77, and J. P. Choi, "Herd Behavior, the 'Penguin Effect,' and Suppression of Informational Diffusion: An Analysis of Informational Externalities and Payoff Interdependency," Rand Journal of Economics, 1997, pp. 407-25.