CHAPTER 26

Clouds

Given extensive and ever-faster connections over both air and fiber optics, the location of computing power relative to the user of said horsepower has become negotiable. Because of the need to do hard things in the service of a smartphone, the awareness of massive inefficiencies in the power consumption of fixed computers, and the cost and complexity of the systems that support a given server, offloading some of the infrastructure to a specialized provider that operates at a huge scale makes sense in ^ ways it did not in the era of the personal computer and local area network.

The proliferation of so-called cloud computing platforms has been rapid. Because there is so much material available that defines the phenomenon (see Figure 26.1), we'll move here to an examination of some of the unexpected consequences and complicated implications of moving some or all of a computing environment to offsite, third-party environments. We will find that the impact of the Internet, as a conduit for both communications and computing, extends far beyond the data center.

To get the problematic and inevitable definitional question out of the way, here is one from Information Week's John Foley: “Cloud computing is on-demand access to virtualized IT resources that are housed outside of your own data center, shared by others, simple to use, paid for via subscription, and accessed over the Web.”¹

There are of course other contending definitions, but Foley's is mercifully brief. Even so, it begs the questions of private clouds, of how small a cloud can be before it starts being something else, and how individual uses of clouds (many people don't own a data center but satisfy most of Foley's other conditions) vary from and overlap corporate ones. It does get us started in more or less the right direction.

FIGURE 26.1 Media Mentions of the Phrase “Cloud Computing,” 2008–2011
Data Source: Google News archives.

Both Technical and Economic Innovation

It's important to note that cloud computing is as much, or more, a matter of economics than of processing cycles: Who invests capital? What products or services are bought and sold or produced and consumed? HP and IBM have long made boxes that created compute capacity; Amazon and Google now sell computing capacity without the box. The end user's business need has not changed—calculating a complex model, recognizing revenue, analyzing transaction histories—but the way resources are organized to meet that need can be radically different.

According to one of the few canonical documents on the topic, a technical report from the University of California at Berkeley, these economic differences are expressed in hardware in three main ways:

1. The illusion of infinite computing resources available on demand, thereby eliminating the need for cloud computing users to plan far ahead for provisioning.
2. The elimination of an up-front commitment by cloud users, thereby allowing companies to start small and increase hardware resources only when there is an increase in their needs.
3. The ability to pay for use of computing resources on a shortterm basis as needed (e.g., processors by the hour and storage by the day) and release them as needed, thereby rewarding conservation by letting machines and storage go when they are no longer useful.²

The rapid pace of change in this area is such that we must add to that list. Marc Benioff is chief executive officer of Salesforce.com, which helped pioneer important aspects of software as a service, a progenitor of modern cloud computing. He further differentiates Amazon, Google, IBM, and other first-generation cloud providers from YouTube, Twitter, and Facebook, all of which, as home to new kinds of applications, serve as cloud providers of a different sort.

As Benioff notes, first-generation clouds were notably low in cost, fast at the compute level, and easy to use. The benefits of the second generation are different: They include collaboration, mobility, and real-time capabilities.³ Thus, the blurring line between hardware (cloud support of lightweight mobile devices such as tablets, e-readers, and smartphones) and people-powered benefits, including incredibly rapid information dissemination and mass collaboration, becomes difficult to trace but important to monitor.

Cloud Computing and the Enterprise

Those large-scale social network cloud behaviors generally are limited to the consumer market, for the moment anyway. Let's return to enterprise computing: As various analysts and technology executives assess the pros and cons of cloud computing, two points of consensus appear to be emerging:

1. Very large data centers benefit from extreme economies of scale.
2. Cloud success stories generally are found outside of the traditional IT shop.

Let us examine each of these in more detail, then probe some of the implications.

Advantages of Scale

Whether run by a cloud provider or a well-managed enterprise IT group, very large data centers exhibit economies of scale not found in smaller server installations. The leverage of relatively expensive and skilled technologists is far higher when one person can manage between 1,000 and 2,000 highly automated servers, as at Microsoft, as opposed to one person being responsible for between 5 and 50 machines, which is common.

In addition, the power consumption of a well-engineered data center can be more efficient than that of many traditional operations. Yahoo! built a new facility in upstate New York, for example, that utilizes atmospheric cooling to the point that only 1% of electricity consumption is for air-conditioning and related cooling tasks.⁴ Having people with deep expertise in cooling, power consumption, recovery, and other niche skills on staff also helps make cloud providers more efficient than those running at smaller scales. The engineering challenges at this scale are themselves new and often fascinating.

Also, large data centers benefit from aggregation of demand. Assume facility A has 10,000 users of computing cycles spread over a variety of different cyclical patterns while facility B has fewer users, all with similar seasonality for retail, quarterly closes for an accounting function, or monthly invoices. Facility A should be able to run more efficiently because it has a more “liquid” market for its capabilities while facility B will likely have to build to its highest load (plus a safety margin) then run less efficiently the majority of the time. What Amazon Distinguished Engineer James Hamilton calls “non-correlated peaks” can be difficult to generate within a single enterprise or function.⁵

IT Organizations Have Yet to Reap the Cloud's Benefits

For all of these benefits, external cloud successes have yet to accrue to traditional IT organizations. At Amazon Web Services, for example, of roughly 200 case studies, none is devoted to traditional enterprise processes such as order management, invoicing and payment processing, or human resources.⁶

There are many readily understandable reasons for this pattern; here is a sample of four. First, legal and regulatory constraints often require a physical audit of information handling practices to which virtual answers are unacceptable. Second, the laws of physics may make large volumes of database joins and other computing tasks difficult to execute off-premise. In general, high-volume transaction processing currently is not recommended as a cloud candidate.

Third, licenses from traditional enterprise providers, such as Microsoft, Oracle, and SAP, are still evolving, making it difficult to run their software in hybrid environments (wherein some processes run locally while others run in a cloud). In addition, only a few enterprise applications of either the package or custom variety are designed to run as well on cloud infrastructure as they do on a conventional server or cluster. Fourth, accounting practices in IT shops may make it difficult to know the true baseline costs and benefits to which an outside provider must compare: some chief information officers never see their electric bills, for example.

The Cloud Will Change How Resources Are Organized

For these reasons, among others, the conclusion is usually drawn that cloud computing is a suboptimal fit for traditional enterprise IT. However, let's invert that logic to see how organizations have historically adapted to new technology capability. When electric motors replaced overhead drive shafts driven by waterwheels adjoining textile mills, the looms and other machines were often left in the same positions for decades before mill owners realized the facility could be organized independently of power supply. More recently, word-processing computers from the likes of Wang initially automated typing pools (one-third of all U.S. women working in 1971 were secretaries); it was not until 10 to 20 years later that large numbers of managers began to service their own document-production needs and thereby alter the shape of organizations.⁷

Enterprise IT architectures embed a wide range of operating assumptions regarding the nature of work, the location of business processes, clock speed, and other factors. When a major shift occurs in the information or other infrastructure, it takes years for organizations to adapt. If we take as our premise that most organizations are not yet prepared to exploit cloud computing (rather than talk about clouds not being ready for “the enterprise”), what are some potential ramifications?

• Organizations are already being founded with very little capital investment. For a services- or knowledge-intensive business that does not make anything physical, free tools and low-cost computing cycles can mostly be expensed, changing the fundraising and indeed organizational strategies significantly.
• The perennial question of “who owns the data?” enters a new phase. While today USB drives and desktop databases continue to make it possible to hoard data, in the future, organizations built on cloud-friendly logic from their origins will deliver new wrinkles to information-handling practices. The issue will by no means disappear: Google's Gmail cloud storage is no doubt already home to a sizable quantity of enterprise data.
• Smartphones, tablets, and other devices built without mass storage can thrive in a cloud-centric environment, particularly if the organization is designed to be fluid and mobile. Coburn Ventures in New York, for example, is an investment firm comprised of a small team of mobile knowledge workers who for the first five years had no corporate office whatsoever: The organization operated from Wi-Fi hotspots, with only occasional all-hands meetings.
• New systems of trust and precautions will need to take shape as the core IT processing capacity migrates to a vendor. It's rarely consequential to contract for a video transcoding or a weather simulation and have it be interrupted. More problematically, near-real-time processes such as customer service likely will need to be redesigned to operate successfully in a cloud or cluster of clouds. Service-level agreements will need to reflect the true cost and impact of interruptions or other lapses. Third-party adjudicators may emerge to assess the responsibility of the cloud customer that introduced a hiccup into the environment relative to the vendor whose failover failed.

Practical Considerations

Because the cloud model introduces innovations in both economics and technology, making educated purchasing and deployment questions requires new metrics, models, and potentially skills.⁸ Some practical questions illustrate the complexity of changing organizational, operational, physical, legal, and computational models on the fly:

1. From the buyer's perspective, what is a vendor's profit path? What can be differentiated and thus generate margins? Compared to the conventional model of data centers, which is often measured in $10,000 or $100,000 increments, cloud computing usage at Amazon is measured in dimes.
2. Related to point 1, how does cloud lock-in vary from existing software (à la classical Microsoft) or hardware (the vintage IBM model) variants? And let's return to that initial Berkeley document: When resources are “released” for other users, where does my data go? If it is in the cloud, the resources can't all be released for other purposes, and if the data moves out of the cloud back to my premise, we both lose cloud advantages (such as power management) and introduce security considerations: Who validates the hard-disk scrub when I retreat from the shared resource?
3. How will incumbents respond? If company A has an established business selling hardware as capital expenditure, and a competing model shifts compute power to an operating-expense model, presumably company B doesn't stand still. How do buyers hedge risk with such a dynamic vendor environment?
4. As with so much of the world's infrastructure, what is the incentive to invest in “pipes” when the value-add lies elsewhere, or nowhere? The robust, high-speed networks upon which the cloud providers rely cannot simply be assumed.
5. If for legal or other reasons the buyer needs performance, security, and/or reliability guarantees, how are these delivered if the buyer cannot see or physically access her assets?
6. There are no free lunches: who bears risk? Every one of the Web's elite destinations has suffered from major outages at some point.⁹ In light of that history, what does a fault-tolerant cloud environment look like, require, and cost? As with so many networked scenarios, the price of failure goes up: When Amazon suffered a cloud outage in 2011, it took dozens of companies without backup down with it.
7. How does optimization work in a cloud? The vendor may be managing to power consumption, say, while customer A wants stable (not necessarily fast, but predictable) transaction times for a shopping-cart scenario. Customer B needs fast compute capability despite big and frequent reads and writes to disc. How can all three parties go home happy at the end of the day?
8. How can virtual, hybrid environments be tested before major real-world events: a quarterly close, a consumer promotion, a currency meltdown? While there will be some pure cloud successes, a big question relates to how well clouds can integrate with existing data centers and other assets. (What constitutes unit testing in a cloud?)
9. What can a customer ask for by way of customization? Who can and will provide it, and at what costs in money and performance? The price points reflect commodity economics, but sooner or later most buyers stumble on needs that surpass plain vanilla.
10. Which standards are open and which are proprietary? The PC architecture flourished in part because of its interoperability: someone could choose a big Maxtor hard drive or a faster Seagate, a Dell flat-panel or Sony Trinitron display, and the hardware maker could buy the cheapest CD drives, memory, and power cords on a given day. USB made the platform more flexible yet. By comparison, once buyers choose a cloud provider, how must they choose an Internet service provider, a system management vendor, a billing system? In short, what are the dependencies introduced by a cloud instance?
11. How fast is fast enough? Cloud computing is a coherent-sounding phrase, but computing has many facets. Think about the different time scales relating to
    • Network latency
    • The laws of physics regarding hard drive access
    • The laws of physics regarding hard drive failure
    • Core competency versus utility workload allocation

    One size clearly cannot fit all.

12. Who guarantees precision? Speaking of laws of physics, all microprocessors are not created equal. Some highly precise calculations, out to many decimal points, might run slightly differently on two different computing cores. How does the user of a given scientific calculation, for example, know that his or her result will be consistent across computing instances, across different clouds, or across brands of processor?

At the end of the day, orchestrating all of those sets of events, each with its own timescape, in a virtual world is a really, really tough technical and managerial problem. Getting the systems to work doesn't even scratch the questions of profitability, liability, audit and related requirements, and so on.

Looking Ahead

For all of those substantial challenges, the question is not whether cloud computing will happen but rather how this tendency will unfold and how organizations, regulators, and other actors will respond. Until the rhetoric and more important the base of experience moves beyond the current state of pilots and vaporware, the range of potential outcomes is too vast to bet on with any serious money. Those problems are soluble. The larger implications are already becoming visible. As cloud computing reallocates the division of labor within the computing fabric, it will also change how managers and, especially, entrepreneurs organize resources into firms, partnerships, and other formal structures. Once these forms emerge, the nature of everything else will be subject to reinvention: work, risk, reward, collaboration, and indeed value itself.

Notes

1. John Foley, “A Definition of Cloud Computing,” InformationWeek Plug Into the Cloud blog, September 26, 2008, www.informationweek.com/cloud-computing/blog/archives/2008/09/a_definition_of.html.

2. Michael Armbrust et al. “Above the Clouds: A Berkeley View of Cloud Computing,” Electrical Engineering and Computer Sciences, University of California at Berkeley technical report # UCB/EECS-2009–28, February 10, 2009, www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.pdf.

3. Marc Benioff, “Welcome to Cloud 2,” Google Latitude conference presentation, April 14, 2010.

4. Rich Miller, “Yahoo Computing Coop: Shape of Things to Come?” Data Center Knowledge, April 26, 2010, www.datacenterknowledge.com/archives/2010/04/26/yahoo-computing-coop-the-shape-of-things-to-come/.

5. James Hamilton, “Private Clouds Are Not the Future,” Perspectives blog, January 17, 2010, http://perspectives.mvdirona.com/2010/01/17/PrivateCloudsAreNotTheFuture.aspx.

6. Amazon Web Services Web site, http://aws.amazon.com/solutions/case-studies/.

7. See also Erik Brynjolfsson, Paul Hofmann, and John Jordan, “Cloud Computing and Electricity: Beyond the Utility Model,” Communications of the ACM 53, no. 5 (May 2010): 32–34, doi 10.1145/1735223.1735234.

8. See “Let It Rise: A Special Report on Corporate IT,” The Economist, October 25, 2008, esp. pp. 13–17, www.economist.com/node/12411882.

9. See, for example, Craig Labovitz, “The Great GoogleLapse,” Arbor Networks Web site, May 14, 2009, http://asert.arbornetworks.com/2009/05/the-great-googlelapse/, and “EBay Apologizes for Web Site Glitch,” CNN Tech, November 23, 2009, http://articles.cnn.com/2009–11-23/tech/ebay.outage_1_ebay-glitch-search?_s=PM:TECH.